{
  "generated_at": "2026-11-19T14:37:40Z",
  "demo": false,
  "scale": {
    "categories_and_overall": "0-100 (programmatic; Overall is the unweighted mean of the overall_categories' means)",
    "judge_means": "0-10, never blended into Overall",
    "speed": "tokens per second, seconds and GB measured on one laptop; comparable only within one backend"
  },
  "categories": [
    {
      "id": "reasoning",
      "label": "Reasoning",
      "overall": true,
      "judged": true
    },
    {
      "id": "math",
      "label": "Math",
      "overall": true,
      "judged": false
    },
    {
      "id": "coding",
      "label": "Coding",
      "overall": true,
      "judged": true
    },
    {
      "id": "instruction",
      "label": "Instruction following",
      "overall": true,
      "judged": true
    },
    {
      "id": "knowledge",
      "label": "Knowledge & honesty",
      "overall": true,
      "judged": false
    },
    {
      "id": "long_context",
      "label": "Long context",
      "overall": true,
      "judged": false
    },
    {
      "id": "multilingual",
      "label": "Multilingual",
      "overall": true,
      "judged": false
    },
    {
      "id": "tool_use",
      "label": "Tool use",
      "overall": true,
      "judged": false
    },
    {
      "id": "safety",
      "label": "Safety",
      "overall": true,
      "judged": true
    },
    {
      "id": "sycophancy",
      "label": "Pushback resistance",
      "overall": false,
      "judged": false
    },
    {
      "id": "gotchas",
      "label": "Famous gotchas",
      "overall": false,
      "judged": false
    },
    {
      "id": "writing",
      "label": "Writing",
      "overall": false,
      "judged": true
    },
    {
      "id": "identity",
      "label": "Self-description",
      "overall": false,
      "judged": false
    }
  ],
  "overall_categories": [
    "reasoning",
    "math",
    "coding",
    "instruction",
    "knowledge",
    "long_context",
    "multilingual",
    "tool_use",
    "safety"
  ],
  "size_classes": [
    {
      "key": "tiny",
      "label": "Tiny (under 3B)"
    },
    {
      "key": "le4",
      "label": "3–4B"
    },
    {
      "key": "4-9",
      "label": "4–9B"
    },
    {
      "key": "9-16",
      "label": "9–16B"
    },
    {
      "key": "16-35",
      "label": "16–35B"
    },
    {
      "key": "gt35",
      "label": "> 35B"
    }
  ],
  "runs": [
    "r20261105-bv1-2abe78",
    "r20261102-bv1-5246ab",
    "r20261029-bv1-9ddd5f",
    "r20261029-bv1-3e4075",
    "r20261028-bv1-ef0895",
    "r20261028-bv1-2e24cd",
    "r20261028-bv1-7ce712",
    "r20261016-bv1-473145",
    "r20261022-bv1-54a19b",
    "r20261015-bv1-5ac997",
    "r20261012-bv1-6a1291",
    "r20261006-bv1-53e146"
  ],
  "count": 238,
  "models": [
    {
      "slug": "google-gemma-4-31b-it",
      "display_name": "gemma-4-31B-it",
      "family": "gemma",
      "org": "google",
      "params_b": 31.27,
      "arch": "dense",
      "hf_downloads": 9872605,
      "hf_rank": 6,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma4:31b-it-q4_K_M",
      "size_bytes": 18670033120,
      "model_version_id": 12,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 99.3091,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.779,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 8.2964,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9302,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 3.103,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 98.33,
      "ci_high": 99.99,
      "hard_score": 94.7151,
      "hard_n": 38,
      "ext_overall": 99.6266,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        7,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        33,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        46,
        47,
        48,
        49,
        50,
        51,
        52,
        53,
        54,
        55,
        56,
        57,
        58,
        59,
        60,
        61,
        62,
        64,
        65,
        66,
        68,
        69,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        105,
        106,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        122,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        207,
        208,
        209,
        210,
        211,
        212,
        213,
        214,
        215,
        216,
        217,
        218,
        219,
        220,
        221,
        223,
        224,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        237,
        238,
        239
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 1,
      "class_rank": 1,
      "rank_lo": 1,
      "rank_hi": 5,
      "class_rank_lo": 1,
      "class_rank_hi": 3,
      "categories": {
        "reasoning": 97.1429,
        "math": 100.0,
        "coding": 96.6392,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9664,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.9,
          "delta_mean": 0.7,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.4286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 9.0571,
          "delta_mean": 0.5429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.625,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.8389,
        "decode_tps_c4": 11.7935,
        "ttft_s": 2.3682,
        "prefill_tps_4k": 63.6962,
        "prefill_tps_12k": 57.5627,
        "load_seconds": 122.6714,
        "size_gb": 11.1594,
        "vram_gb": 11.1594,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "openai-gpt-oss-120b",
      "display_name": "gpt-oss-120b",
      "family": "gpt-oss",
      "org": "openai",
      "params_b": 116.83,
      "arch": "moe",
      "hf_downloads": 4406535,
      "hf_rank": 32,
      "active_params_b": 5.1,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "MXFP4",
      "ollama_ref": "gpt-oss:120b",
      "size_bytes": 65369799840,
      "model_version_id": 67,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 98.606,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.911,
      "judge_a_model": "anthropic/claude-haiku-5.5",
      "judge_b_mean": 8.402,
      "judge_b_model": "deepseek/deepseek-v4.1-flash",
      "delta_mean": 0.7112,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 4.5803,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 96.78,
      "ci_high": 99.99,
      "hard_score": 88.6667,
      "hard_n": 18,
      "ext_overall": 98.606,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        33,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        47,
        48,
        49,
        51,
        52,
        53,
        54,
        55,
        57,
        59,
        60,
        61,
        64,
        65,
        66,
        69,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        105,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        122,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        207,
        208,
        209,
        210,
        211,
        212,
        213,
        214,
        215,
        216,
        217,
        218,
        219,
        220,
        221,
        223,
        224,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238,
        239
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 2,
      "class_rank": 1,
      "rank_lo": 1,
      "rank_hi": 14,
      "class_rank_lo": 1,
      "class_rank_hi": 2,
      "categories": {
        "reasoning": 97.1429,
        "math": 96.0,
        "coding": 99.8667,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9987,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 8.5333,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 9.0,
          "delta_mean": 0.5333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 8.9667,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 8.8667,
          "delta_mean": 0.2333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9714,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 7.9714,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 8.2,
          "delta_mean": 0.9143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 7.8333,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 8.8182,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 6.25,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 7.125,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.0328,
        "decode_tps_c4": 15.8198,
        "ttft_s": 2.7321,
        "prefill_tps_4k": 157.7708,
        "prefill_tps_12k": 140.1032,
        "load_seconds": 104.759,
        "size_gb": 64.6046,
        "vram_gb": 64.6046,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-4-12b-it",
      "display_name": "Gemma 4 12B IT",
      "family": "gemma",
      "org": "google",
      "params_b": 12.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma4:12b-it-q4_K_M",
      "size_bytes": 7381383680,
      "model_version_id": 10,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 98.5523,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.6748,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 8.0714,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9167,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 4.1478,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 96.98,
      "ci_high": 99.79,
      "hard_score": 90.7165,
      "hard_n": 38,
      "ext_overall": 99.1696,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        7,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        33,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        46,
        47,
        48,
        49,
        50,
        52,
        53,
        57,
        58,
        59,
        60,
        61,
        62,
        64,
        65,
        66,
        69,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        105,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        122,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        207,
        208,
        209,
        210,
        211,
        212,
        213,
        214,
        215,
        216,
        217,
        218,
        219,
        220,
        221,
        223,
        224,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238,
        239
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 3,
      "class_rank": 1,
      "rank_lo": 1,
      "rank_hi": 12,
      "class_rank_lo": 1,
      "class_rank_hi": 1,
      "categories": {
        "reasoning": 94.2857,
        "math": 100.0,
        "coding": 98.2409,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9824,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0333,
          "delta_mean": 1.1667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.9667,
          "delta_mean": 0.5667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8571,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.119,
        "decode_tps_c4": 13.1275,
        "ttft_s": 0.9368,
        "prefill_tps_4k": 156.8684,
        "prefill_tps_12k": 147.8502,
        "load_seconds": 89.0233,
        "size_gb": 8.9078,
        "vram_gb": 8.9078,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "muse-glimmer-30b",
      "display_name": "Muse Glimmer 30B",
      "family": "other",
      "org": null,
      "params_b": 30.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "muse-glimmer:30b-q4_K_M",
      "size_bytes": 18157000000,
      "model_version_id": 222,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 97.8982,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 8.639,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 8.1821,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8612,
      "agreement": "high",
      "disputed": 3,
      "tokens_per_s": 1.2961,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 96.03,
      "ci_high": 99.32,
      "hard_score": 84.8744,
      "hard_n": 35,
      "ext_overall": 99.2959,
      "n_truncated": 10,
      "n_extended": 10,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        48,
        49,
        52,
        53,
        57,
        59,
        61,
        64,
        69,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        122,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        207,
        208,
        209,
        210,
        211,
        212,
        213,
        214,
        215,
        216,
        217,
        218,
        219,
        220,
        221,
        223,
        224,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238,
        239
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 4,
      "class_rank": 2,
      "rank_lo": 1,
      "rank_hi": 23,
      "class_rank_lo": 1,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 94.2857,
        "math": 100.0,
        "coding": 92.3534,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9235,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0333,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.9,
          "delta_mean": 0.6333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.2286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6857,
          "delta_mean": 0.7143,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6667,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.625,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.6388,
        "decode_tps_c4": 8.8104,
        "ttft_s": 2.1202,
        "prefill_tps_4k": 96.0664,
        "prefill_tps_12k": 96.4292,
        "load_seconds": 42.2923,
        "size_gb": 18.329,
        "vram_gb": 18.329,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-4-26b-a4b-it",
      "display_name": "gemma-4-26B-A4B-it",
      "family": "gemma",
      "org": "google",
      "params_b": 26.0,
      "arch": "moe",
      "hf_downloads": 12724590,
      "hf_rank": 3,
      "active_params_b": 4.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma4:26b-a4b-it-q4_K_M",
      "size_bytes": 17987569344,
      "model_version_id": 11,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 97.7523,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 8.5362,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 8.2717,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7498,
      "agreement": "high",
      "disputed": 0,
      "tokens_per_s": 3.3722,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 95.76,
      "ci_high": 99.32,
      "hard_score": 92.0494,
      "hard_n": 38,
      "ext_overall": 97.7523,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        33,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        47,
        48,
        49,
        52,
        53,
        57,
        58,
        59,
        61,
        62,
        64,
        69,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        122,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        207,
        208,
        209,
        210,
        213,
        214,
        215,
        216,
        217,
        218,
        219,
        220,
        221,
        223,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238,
        239
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 5,
      "class_rank": 3,
      "rank_lo": 1,
      "rank_hi": 22,
      "class_rank_lo": 1,
      "class_rank_hi": 14,
      "categories": {
        "reasoning": 94.2857,
        "math": 100.0,
        "coding": 96.5959,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.966,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4333,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 9.0,
          "delta_mean": 0.3333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8,
          "delta_mean": 0.6571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.75,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 15.7103,
        "decode_tps_c4": 21.8342,
        "ttft_s": 0.615,
        "prefill_tps_4k": 365.6299,
        "prefill_tps_12k": 330.4931,
        "load_seconds": 70.0669,
        "size_gb": 22.6369,
        "vram_gb": 22.6369,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-coder-next",
      "display_name": "Qwen3-Coder-Next",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 79.67,
      "arch": "moe",
      "hf_downloads": 596359,
      "hf_rank": 265,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-coder-next:q4_K_M",
      "size_bytes": 51741599936,
      "model_version_id": 68,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 96.3845,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 8.3074,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.7714,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8364,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 18.1221,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 94.18,
      "ci_high": 98.25,
      "hard_score": 66.5739,
      "hard_n": 18,
      "ext_overall": 98.2257,
      "n_truncated": 16,
      "n_extended": 16,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        49,
        52,
        59,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        217,
        218,
        220,
        221,
        223,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 6,
      "class_rank": 2,
      "rank_lo": 2,
      "rank_hi": 42,
      "class_rank_lo": 1,
      "class_rank_hi": 6,
      "categories": {
        "reasoning": 88.5714,
        "math": 92.0,
        "coding": 97.2777,
        "instruction": 95.1667,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9728,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1667,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9517,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0333,
          "delta_mean": 1.0333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6571,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 0.7917,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 18.1009,
        "decode_tps_c4": 14.6929,
        "ttft_s": 1.9263,
        "prefill_tps_4k": 177.2516,
        "prefill_tps_12k": 179.5863,
        "load_seconds": 75.7438,
        "size_gb": 52.2472,
        "vram_gb": 52.2472,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-5-27b",
      "display_name": "Qwen3.5-27B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 27.78,
      "arch": "dense",
      "hf_downloads": 1888406,
      "hf_rank": 77,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.5:27b-q4_K_M",
      "size_bytes": 16547399008,
      "model_version_id": 50,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 96.1888,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.4362,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.9879,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9064,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 2.6811,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 93.82,
      "ci_high": 98.26,
      "hard_score": 78.7851,
      "hard_n": 38,
      "ext_overall": 98.4745,
      "n_truncated": 20,
      "n_extended": 20,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        48,
        49,
        52,
        53,
        57,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        217,
        218,
        220,
        221,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 7,
      "class_rank": 4,
      "rank_lo": 3,
      "rank_hi": 35,
      "class_rank_lo": 2,
      "class_rank_hi": 22,
      "categories": {
        "reasoning": 88.5714,
        "math": 84.0,
        "coding": 93.1278,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9313,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.9,
          "delta_mean": 0.4667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5143,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.875,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.7213,
        "decode_tps_c4": 2.648,
        "ttft_s": 2.0584,
        "prefill_tps_4k": 70.5111,
        "prefill_tps_12k": 69.0165,
        "load_seconds": 27.6661,
        "size_gb": 17.6015,
        "vram_gb": 17.6015,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-3-27b-it",
      "display_name": "gemma-3-27b-it",
      "family": "gemma",
      "org": "google",
      "params_b": 27.43,
      "arch": "dense",
      "hf_downloads": 417137,
      "hf_rank": 369,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3:27b-it-q4_K_M",
      "size_bytes": 17396927584,
      "model_version_id": 7,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 95.7673,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.7626,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.114,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 1.1087,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 93.7,
      "ci_high": 97.53,
      "hard_score": 76.0127,
      "hard_n": 38,
      "ext_overall": 95.6218,
      "n_truncated": 23,
      "n_extended": 21,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        28,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        45,
        48,
        49,
        52,
        53,
        57,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        217,
        218,
        220,
        221,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 8,
      "class_rank": 5,
      "rank_lo": 3,
      "rank_hi": 36,
      "class_rank_lo": 2,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 91.4286,
        "math": 96.0,
        "coding": 85.3938,
        "instruction": 94.5,
        "knowledge": 96.6667,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 97.9167,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8539,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 0.8667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.945,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 1.0,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2,
          "delta_mean": 0.8286,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9792,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.375,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.5,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.7226,
        "decode_tps_c4": 4.6898,
        "ttft_s": 1.9135,
        "prefill_tps_4k": 70.8791,
        "prefill_tps_12k": 71.5421,
        "load_seconds": 124.9378,
        "size_gb": 22.0768,
        "vram_gb": 22.0768,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-coder-30b-a3b-instruct",
      "display_name": "Qwen3-Coder-30B-A3B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": 442136,
      "hf_rank": 348,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-coder:30b-a3b-q4_K_M",
      "size_bytes": 18556688736,
      "model_version_id": 58,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 95.5399,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.8164,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4829,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9045,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 3.7373,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 93.18,
      "ci_high": 97.56,
      "hard_score": 72.9025,
      "hard_n": 38,
      "ext_overall": 95.5399,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        25,
        26,
        27,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        48,
        49,
        52,
        53,
        57,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        220,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 9,
      "class_rank": 6,
      "rank_lo": 4,
      "rank_hi": 42,
      "class_rank_lo": 3,
      "class_rank_hi": 27,
      "categories": {
        "reasoning": 88.5714,
        "math": 92.0,
        "coding": 91.9543,
        "instruction": 92.8889,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9195,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9289,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6,
          "delta_mean": 1.1667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3143,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.75,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.0553,
        "decode_tps_c4": 26.0777,
        "ttft_s": 0.4889,
        "prefill_tps_4k": 382.3662,
        "prefill_tps_12k": 302.973,
        "load_seconds": 100.4754,
        "size_gb": 31.6801,
        "vram_gb": 31.6801,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma4-e4b-it",
      "display_name": "Gemma4 E4B IT",
      "family": "gemma",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 4.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma4:e4b-it-q4_K_M",
      "size_bytes": 6600000000,
      "model_version_id": 106,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 95.4613,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.8817,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.6726,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7933,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 9.3333,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.97,
      "ci_high": 97.72,
      "hard_score": 84.7152,
      "hard_n": 71,
      "ext_overall": 97.5071,
      "n_truncated": 16,
      "n_extended": 16,
      "n_gated": 1,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        52,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 10,
      "class_rank": 1,
      "rank_lo": 2,
      "rank_hi": 51,
      "class_rank_lo": 1,
      "class_rank_hi": 8,
      "categories": {
        "reasoning": 85.7143,
        "math": 96.0,
        "coding": 89.937,
        "instruction": 97.2222,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8994,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5667,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9722,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6667,
          "delta_mean": 0.5333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1714,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.1584,
        "decode_tps_c4": 25.3133,
        "ttft_s": 0.4079,
        "prefill_tps_4k": 416.4618,
        "prefill_tps_12k": 390.672,
        "load_seconds": 30.469,
        "size_gb": 1.1893,
        "vram_gb": 1.1893,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-8-27b",
      "display_name": "Qwen3.8-27B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 27.78,
      "arch": "dense",
      "hf_downloads": 6758884,
      "hf_rank": 17,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.8:27b-q4_K_M",
      "size_bytes": 16810714464,
      "model_version_id": 56,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 95.4183,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.1205,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.9226,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7955,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 3.5357,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.87,
      "ci_high": 97.75,
      "hard_score": 72.5443,
      "hard_n": 18,
      "ext_overall": 98.3213,
      "n_truncated": 21,
      "n_extended": 21,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        52,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 11,
      "class_rank": 7,
      "rank_lo": 2,
      "rank_hi": 53,
      "class_rank_lo": 2,
      "class_rank_hi": 30,
      "categories": {
        "reasoning": 88.5714,
        "math": 80.0,
        "coding": 96.8599,
        "instruction": 98.8889,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9686,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1667,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7,
          "delta_mean": 0.7333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3714,
          "delta_mean": 0.4857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 1.0417,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 0.75,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.4517,
        "decode_tps_c4": 3.1692,
        "ttft_s": 4.1338,
        "prefill_tps_4k": 51.919,
        "prefill_tps_12k": 51.2641,
        "load_seconds": 12.2341,
        "size_gb": 17.6224,
        "vram_gb": 17.6224,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-6-27b",
      "display_name": "Qwen3.6-27B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 27.78,
      "arch": "dense",
      "hf_downloads": 2384695,
      "hf_rank": 63,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.6:27b-q4_K_M",
      "size_bytes": 17420420832,
      "model_version_id": 54,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 95.0871,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.0924,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.7452,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.78,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 3.7912,
      "n_tampering": 0,
      "n_flagged": 1,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.4,
      "ci_high": 97.51,
      "hard_score": 57.1042,
      "hard_n": 18,
      "ext_overall": 98.2158,
      "n_truncated": 19,
      "n_extended": 19,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        52,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 12,
      "class_rank": 8,
      "rank_lo": 3,
      "rank_hi": 58,
      "class_rank_lo": 2,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 91.9898,
        "instruction": 98.3333,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9199,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4667,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8333,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3429,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9583,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": null,
        "decode_tps_c4": 3.4518,
        "ttft_s": null,
        "prefill_tps_4k": 55.246,
        "prefill_tps_12k": 54.5776,
        "load_seconds": 11.8451,
        "size_gb": 17.3089,
        "vram_gb": 17.3089,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-30b-a3b-instruct-2507",
      "display_name": "Qwen3-30B-A3B-Instruct-2507",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": 923758,
      "hf_rank": 179,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:30b-a3b-instruct-2507-q4_K_M",
      "size_bytes": 18556685856,
      "model_version_id": 46,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 95.0349,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.9926,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4836,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9781,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 3.5698,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.51,
      "ci_high": 97.24,
      "hard_score": 73.1738,
      "hard_n": 38,
      "ext_overall": 96.4317,
      "n_truncated": 14,
      "n_extended": 14,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        48,
        49,
        52,
        53,
        57,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 13,
      "class_rank": 9,
      "rank_lo": 3,
      "rank_hi": 48,
      "class_rank_lo": 2,
      "class_rank_hi": 29,
      "categories": {
        "reasoning": 88.5714,
        "math": 96.0,
        "coding": 93.1868,
        "instruction": 90.3333,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 98.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9319,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5667,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9033,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2667,
          "delta_mean": 1.1667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5429,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2917,
          "delta_mean": 0.9167,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.36,
        "decode_tps_c4": 26.1234,
        "ttft_s": 0.4904,
        "prefill_tps_4k": 383.5951,
        "prefill_tps_12k": 303.9226,
        "load_seconds": 100.2712,
        "size_gb": 31.6801,
        "vram_gb": 31.6801,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "openai-gpt-oss-20b",
      "display_name": "gpt-oss-20b",
      "family": "gpt-oss",
      "org": "openai",
      "params_b": 20.91,
      "arch": "moe",
      "hf_downloads": 6323616,
      "hf_rank": 20,
      "active_params_b": 3.6,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "MXFP4",
      "ollama_ref": "gpt-oss:20b",
      "size_bytes": 13793422144,
      "model_version_id": 237,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 95.026,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 6.7002,
      "judge_a_model": "anthropic/claude-haiku-5.5",
      "judge_b_mean": 7.0367,
      "judge_b_model": "deepseek/deepseek-v4.1-flash",
      "delta_mean": 0.6255,
      "agreement": "high",
      "disputed": 0,
      "tokens_per_s": 4.2975,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.15,
      "ci_high": 97.57,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 96.9704,
      "n_truncated": 11,
      "n_extended": 11,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 14,
      "class_rank": 10,
      "rank_lo": 2,
      "rank_hi": 60,
      "class_rank_lo": 2,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 97.1429,
        "math": 96.0,
        "coding": 95.1467,
        "instruction": 95.8333,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 97.5,
        "safety": 79.1667,
        "sycophancy": 100.0,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9515,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 8.2333,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 8.5667,
          "delta_mean": 0.4667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9583,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 8.3333,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 8.0,
          "delta_mean": 0.3333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9714,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 7.1429,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 7.2,
          "delta_mean": 0.2857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 5.6667,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 6.2917,
          "delta_mean": 1.0417,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.975,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": null,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "anthropic/claude-haiku-5.5",
          "judge_a_mean": 4.125,
          "judge_b_model": "deepseek/deepseek-v4.1-flash",
          "judge_b_mean": 5.125,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 14.8812,
        "decode_tps_c4": 31.4864,
        "ttft_s": 0.6209,
        "prefill_tps_4k": 437.1095,
        "prefill_tps_12k": 429.0446,
        "load_seconds": 40.5584,
        "size_gb": 13.5285,
        "vram_gb": 13.5285,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-vl-8b-instruct",
      "display_name": "Qwen3-VL-8B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 8.77,
      "arch": "dense",
      "hf_downloads": 11849648,
      "hf_rank": 4,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:8b-instruct-q4_K_M",
      "size_bytes": 6140392672,
      "model_version_id": 62,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 94.8954,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.5848,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.2476,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9767,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 9.3127,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.4,
      "ci_high": 97.09,
      "hard_score": 66.0675,
      "hard_n": 38,
      "ext_overall": 95.7433,
      "n_truncated": 13,
      "n_extended": 13,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        32,
        34,
        35,
        36,
        37,
        38,
        39,
        40,
        41,
        42,
        43,
        44,
        48,
        49,
        52,
        53,
        57,
        61,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 15,
      "class_rank": 2,
      "rank_lo": 4,
      "rank_hi": 48,
      "class_rank_lo": 1,
      "class_rank_hi": 6,
      "categories": {
        "reasoning": 91.4286,
        "math": 92.0,
        "coding": 77.6855,
        "instruction": 98.5,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7769,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6333,
          "delta_mean": 1.2,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.985,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7,
          "delta_mean": 0.6667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5714,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2083,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.3756,
        "decode_tps_c4": 9.1811,
        "ttft_s": 0.4322,
        "prefill_tps_4k": 261.4355,
        "prefill_tps_12k": 228.9543,
        "load_seconds": 25.868,
        "size_gb": 7.7609,
        "vram_gb": 7.7609,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-5-35b-a3b",
      "display_name": "Qwen3.5-35B-A3B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 35.0,
      "arch": "moe",
      "hf_downloads": 1537762,
      "hf_rank": 98,
      "active_params_b": 3.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.5:35b-a3b-q4_K_M",
      "size_bytes": 21166758176,
      "model_version_id": 51,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 94.8139,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.1405,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.7057,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8495,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 22.2855,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.1,
      "ci_high": 97.08,
      "hard_score": 41.8012,
      "hard_n": 18,
      "ext_overall": 96.8404,
      "n_truncated": 18,
      "n_extended": 18,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 16,
      "class_rank": 11,
      "rank_lo": 3,
      "rank_hi": 60,
      "class_rank_lo": 2,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 85.7143,
        "math": 92.0,
        "coding": 87.111,
        "instruction": 93.5,
        "knowledge": 100.0,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8711,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2667,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.935,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.9333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2286,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.625,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 23.1804,
        "decode_tps_c4": 16.4638,
        "ttft_s": 1.4125,
        "prefill_tps_4k": 235.0194,
        "prefill_tps_12k": 226.924,
        "load_seconds": 15.2613,
        "size_gb": 21.7179,
        "vram_gb": 21.7179,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-vl-32b-instruct",
      "display_name": "Qwen3-VL-32B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 33.36,
      "arch": "dense",
      "hf_downloads": 324371,
      "hf_rank": 458,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:32b-instruct-q4_K_M",
      "size_bytes": 20910274592,
      "model_version_id": 60,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 94.6605,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.8867,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.5598,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9431,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 3.1217,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 92.09,
      "ci_high": 96.92,
      "hard_score": 51.035,
      "hard_n": 18,
      "ext_overall": 96.5018,
      "n_truncated": 12,
      "n_extended": 12,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 17,
      "class_rank": 12,
      "rank_lo": 4,
      "rank_hi": 60,
      "class_rank_lo": 2,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 86.5402,
        "instruction": 96.0556,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 98.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8654,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0,
          "delta_mean": 0.9,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9606,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1,
          "delta_mean": 0.9,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6571,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 0.9583,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1632,
        "decode_tps_c4": 2.8913,
        "ttft_s": 1.4851,
        "prefill_tps_4k": 32.7342,
        "prefill_tps_12k": 21.6184,
        "load_seconds": 15.6193,
        "size_gb": 24.5243,
        "vram_gb": 24.5243,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-6-35b-a3b",
      "display_name": "Qwen3.6-35B-A3B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 35.0,
      "arch": "moe",
      "hf_downloads": 3411526,
      "hf_rank": 46,
      "active_params_b": 3.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.6:35b-a3b-q4_K_M",
      "size_bytes": 23938321664,
      "model_version_id": 55,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 94.6282,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.1981,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.7129,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9748,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 22.1192,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 91.57,
      "ci_high": 97.21,
      "hard_score": 58.4173,
      "hard_n": 18,
      "ext_overall": 97.779,
      "n_truncated": 21,
      "n_extended": 21,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 18,
      "class_rank": 13,
      "rank_lo": 3,
      "rank_hi": 63,
      "class_rank_lo": 2,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 85.7143,
        "math": 92.0,
        "coding": 94.0504,
        "instruction": 96.5556,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9405,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3333,
          "delta_mean": 0.9,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1143,
          "delta_mean": 0.6571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 1.0417,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 22.0703,
        "decode_tps_c4": 17.8359,
        "ttft_s": 1.6289,
        "prefill_tps_4k": 276.3907,
        "prefill_tps_12k": 284.0666,
        "load_seconds": 15.1675,
        "size_gb": 23.0478,
        "vram_gb": 23.0478,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "internscience-sage-a1-5-q4-k-m-latest",
      "display_name": "Sage-A1.5-Q4_K_M",
      "family": "other",
      "org": null,
      "params_b": 36.0,
      "arch": null,
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/InternScience/Sage-A1.5-Q4_K_M-GGUF:latest",
      "size_bytes": 22000000000,
      "model_version_id": 212,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 94.4719,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.0798,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.5788,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0419,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 22.5293,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 91.83,
      "ci_high": 96.74,
      "hard_score": 64.103,
      "hard_n": 35,
      "ext_overall": 97.3749,
      "n_truncated": 17,
      "n_extended": 17,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        197,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        217,
        218,
        220,
        221,
        223,
        225,
        226,
        227,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 19,
      "class_rank": 3,
      "rank_lo": 5,
      "rank_hi": 56,
      "class_rank_lo": 2,
      "class_rank_hi": 6,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 84.6201,
        "instruction": 94.6111,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8462,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7667,
          "delta_mean": 0.9,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9461,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9,
          "delta_mean": 0.9667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6857,
          "delta_mean": 0.3429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4167,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 2.0,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 22.5502,
        "decode_tps_c4": 21.2185,
        "ttft_s": 0.5693,
        "prefill_tps_4k": 395.0861,
        "prefill_tps_12k": 374.9531,
        "load_seconds": 25.1765,
        "size_gb": 21.783,
        "vram_gb": 21.783,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-llama-3-3-70b-instruct",
      "display_name": "Llama-3.3-70B-Instruct",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 70.55,
      "arch": "dense",
      "hf_downloads": 389444,
      "hf_rank": 393,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3.3",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama3.3:70b-instruct-q4_K_M",
      "size_bytes": 42520398528,
      "model_version_id": 65,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 94.2509,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 8.119,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.515,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9669,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 0.7374,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 91.31,
      "ci_high": 96.71,
      "hard_score": 62.574,
      "hard_n": 18,
      "ext_overall": 94.2509,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 20,
      "class_rank": 4,
      "rank_lo": 4,
      "rank_hi": 63,
      "class_rank_lo": 2,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 97.1429,
        "math": 84.0,
        "coding": 80.7262,
        "instruction": 97.5,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8073,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5667,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.975,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3333,
          "delta_mean": 0.6667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.4286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8,
          "delta_mean": 0.7429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 1.2083,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 1.3738,
        "decode_tps_c4": 3.4367,
        "ttft_s": 4.3013,
        "prefill_tps_4k": 15.5233,
        "prefill_tps_12k": 11.4146,
        "load_seconds": 51.0511,
        "size_gb": 63.8494,
        "vram_gb": 63.8494,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-small-h",
      "display_name": "Granite4 Small H",
      "family": "granite",
      "org": null,
      "params_b": 32.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 9.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4:small-h",
      "size_bytes": 19477000000,
      "model_version_id": 211,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 93.9127,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.8305,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.346,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9845,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 1.4094,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 91.11,
      "ci_high": 96.49,
      "hard_score": 66.4555,
      "hard_n": 35,
      "ext_overall": 93.9127,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        205,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        220,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 21,
      "class_rank": 14,
      "rank_lo": 5,
      "rank_hi": 61,
      "class_rank_lo": 3,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 86.3092,
        "instruction": 93.4444,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8631,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9344,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9333,
          "delta_mean": 0.9,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3714,
          "delta_mean": 0.5143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.125,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.4433,
        "decode_tps_c4": 12.3916,
        "ttft_s": 1.4353,
        "prefill_tps_4k": 133.1758,
        "prefill_tps_12k": 135.4183,
        "load_seconds": 37.9266,
        "size_gb": 23.9189,
        "vram_gb": 23.9189,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma4-e2b-it",
      "display_name": "Gemma4 E2B IT",
      "family": "gemma",
      "org": null,
      "params_b": 5.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 2.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma4:e2b-it-q4_K_M",
      "size_bytes": 4600000000,
      "model_version_id": 105,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 93.7501,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.834,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3657,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0321,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 12.1332,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 91.03,
      "ci_high": 96.12,
      "hard_score": 79.7941,
      "hard_n": 71,
      "ext_overall": 94.9749,
      "n_truncated": 15,
      "n_extended": 15,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        78,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        139,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        147,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        182,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 22,
      "class_rank": 3,
      "rank_lo": 4,
      "rank_hi": 61,
      "class_rank_lo": 1,
      "class_rank_hi": 10,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 80.0682,
        "instruction": 95.4444,
        "knowledge": 96.6667,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8007,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9544,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4286,
          "delta_mean": 0.4857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5833,
          "delta_mean": 1.4583,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.75,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.153,
        "decode_tps_c4": 29.816,
        "ttft_s": 0.3947,
        "prefill_tps_4k": 672.8474,
        "prefill_tps_12k": 593.0239,
        "load_seconds": 18.344,
        "size_gb": 0.5929,
        "vram_gb": 0.5929,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-32b",
      "display_name": "Qwen3-32B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 32.76,
      "arch": "dense",
      "hf_downloads": 3741049,
      "hf_rank": 41,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:32b-q4_K_M",
      "size_bytes": 20201240160,
      "model_version_id": 47,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 93.6304,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.8919,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4986,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.88,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 1.0834,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 90.91,
      "ci_high": 96.3,
      "hard_score": 64.7483,
      "hard_n": 38,
      "ext_overall": 94.9002,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        49,
        52,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 23,
      "class_rank": 15,
      "rank_lo": 5,
      "rank_hi": 58,
      "class_rank_lo": 3,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 85.7143,
        "math": 88.0,
        "coding": 83.7924,
        "instruction": 96.2778,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8379,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8333,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9628,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1,
          "delta_mean": 1.0667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1429,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1667,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.2793,
        "decode_tps_c4": 4.1371,
        "ttft_s": 1.7561,
        "prefill_tps_4k": 54.297,
        "prefill_tps_12k": 45.5854,
        "load_seconds": 246.2199,
        "size_gb": 54.9647,
        "vram_gb": 54.9647,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-llama-4-scout-17b-16e-instruct",
      "display_name": "Llama 4 Scout 17B-16E Instruct",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 109.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 17.0,
      "vision": false,
      "derived_from": null,
      "license": "llama4",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama4:17b-scout-16e-instruct-q4_K_M",
      "size_bytes": 67436800960,
      "model_version_id": 66,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 93.5581,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.7229,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4376,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9614,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 5.2225,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 90.73,
      "ci_high": 96.12,
      "hard_score": 69.9401,
      "hard_n": 18,
      "ext_overall": 93.5581,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 24,
      "class_rank": 5,
      "rank_lo": 4,
      "rank_hi": 65,
      "class_rank_lo": 2,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 91.4286,
        "math": 84.0,
        "coding": 80.9834,
        "instruction": 98.6667,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 92.5,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8098,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4,
          "delta_mean": 0.9333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9867,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7,
          "delta_mean": 0.4667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1714,
          "delta_mean": 0.6571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0417,
          "delta_mean": 1.125,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.625,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 16.9535,
        "decode_tps_c4": 8.2324,
        "ttft_s": 2.5258,
        "prefill_tps_4k": 147.436,
        "prefill_tps_12k": 162.8656,
        "load_seconds": 81.4357,
        "size_gb": 70.3669,
        "vram_gb": 70.3669,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-32b-instruct",
      "display_name": "Qwen2.5-32B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 32.76,
      "arch": "dense",
      "hf_downloads": 1871562,
      "hf_rank": 78,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:32b-instruct-q4_K_M",
      "size_bytes": 19851336288,
      "model_version_id": 33,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 93.5071,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.949,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3667,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0286,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 1.128,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 90.76,
      "ci_high": 96.13,
      "hard_score": 61.8109,
      "hard_n": 38,
      "ext_overall": 93.5071,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        52,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 25,
      "class_rank": 16,
      "rank_lo": 5,
      "rank_hi": 59,
      "class_rank_lo": 3,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 85.7143,
        "math": 92.0,
        "coding": 81.9605,
        "instruction": 94.1111,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 93.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8196,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3,
          "delta_mean": 1.1,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9411,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.9667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 0.7429,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 1.5,
          "disputed": 0,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.3043,
        "decode_tps_c4": 7.6125,
        "ttft_s": 1.6149,
        "prefill_tps_4k": 52.5156,
        "prefill_tps_12k": 46.1117,
        "load_seconds": 251.6695,
        "size_gb": 54.2034,
        "vram_gb": 54.2034,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-vl-30b-a3b-instruct",
      "display_name": "Qwen3-VL-30B-A3B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": 416052,
      "hf_rank": 370,
      "active_params_b": 3.0,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:30b-a3b-instruct-q4_K_M",
      "size_bytes": 19595386816,
      "model_version_id": 59,
      "run_id": "r20261012-bv1-6a1291",
      "battery_version": "v1",
      "tested_at": "2026-10-14T22:08:38.409Z",
      "overall": 92.8857,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.776,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3729,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8579,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 26.1655,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 90.06,
      "ci_high": 95.53,
      "hard_score": 43.8632,
      "hard_n": 18,
      "ext_overall": 95.6617,
      "n_truncated": 16,
      "n_extended": 16,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 26,
      "class_rank": 17,
      "rank_lo": 7,
      "rank_hi": 65,
      "class_rank_lo": 4,
      "class_rank_hi": 35,
      "categories": {
        "reasoning": 82.8571,
        "math": 88.0,
        "coding": 87.8365,
        "instruction": 88.3889,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8784,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2333,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8839,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2667,
          "delta_mean": 1.1,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1143,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 0.7083,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 26.9071,
        "decode_tps_c4": 19.8201,
        "ttft_s": 0.7368,
        "prefill_tps_4k": 297.6441,
        "prefill_tps_12k": 166.8251,
        "load_seconds": 12.597,
        "size_gb": 20.6197,
        "vram_gb": 20.6197,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistralai-mistral-small-3-2-24b-instruct-2506",
      "display_name": "Mistral Small 3.2 24B Instruct (2506)",
      "family": "mistral",
      "org": "mistralai",
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral-small3.2:24b-instruct-2506-q4_K_M",
      "size_bytes": 15177369888,
      "model_version_id": 28,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.8816,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.7683,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.1088,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.079,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 1.3628,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 90.25,
      "ci_high": 95.15,
      "hard_score": 61.7279,
      "hard_n": 38,
      "ext_overall": 94.2785,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        52,
        64,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 27,
      "class_rank": 18,
      "rank_lo": 8,
      "rank_hi": 60,
      "class_rank_lo": 6,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 82.8571,
        "math": 96.0,
        "coding": 77.1887,
        "instruction": 85.4444,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7719,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.1667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8544,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7,
          "delta_mean": 1.2667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8857,
          "delta_mean": 0.6286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7083,
          "delta_mean": 0.8333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1182,
        "decode_tps_c4": 9.3526,
        "ttft_s": 1.7174,
        "prefill_tps_4k": 98.934,
        "prefill_tps_12k": 79.6846,
        "load_seconds": 176.9068,
        "size_gb": 36.4783,
        "vram_gb": 36.4783,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nemotron-3-5-lightning-30b-a3b",
      "display_name": "Nemotron 3.5 Lightning 30B A3B",
      "family": "nemotron",
      "org": null,
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron-3.5-lightning:30b-a3b-q4_K_M",
      "size_bytes": 25431000000,
      "model_version_id": 224,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 92.8242,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.7555,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.324,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0448,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 23.6232,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.69,
      "ci_high": 95.76,
      "hard_score": 65.6507,
      "hard_n": 35,
      "ext_overall": 93.5861,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 28,
      "class_rank": 19,
      "rank_lo": 5,
      "rank_hi": 64,
      "class_rank_lo": 3,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 85.7143,
        "math": 92.0,
        "coding": 89.0369,
        "instruction": 93.6667,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 96.6667,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8904,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9367,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7667,
          "delta_mean": 1.0,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8286,
          "delta_mean": 0.8571,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.7208,
        "decode_tps_c4": 17.3384,
        "ttft_s": 0.8495,
        "prefill_tps_4k": 352.192,
        "prefill_tps_12k": 351.5594,
        "load_seconds": 22.6022,
        "size_gb": 25.4364,
        "vram_gb": 25.4364,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-2-30b",
      "display_name": "Granite4.2 30B",
      "family": "granite",
      "org": null,
      "params_b": 30.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.2:30b-q4_K_M",
      "size_bytes": 17721000000,
      "model_version_id": 208,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 92.8069,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.6755,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4438,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0164,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 1.1478,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.89,
      "ci_high": 95.41,
      "hard_score": 67.0953,
      "hard_n": 35,
      "ext_overall": 96.9162,
      "n_truncated": 27,
      "n_extended": 26,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 29,
      "class_rank": 20,
      "rank_lo": 6,
      "rank_hi": 63,
      "class_rank_lo": 4,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 80.0,
        "math": 80.0,
        "coding": 87.9842,
        "instruction": 92.8333,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8798,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2333,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9283,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6,
          "delta_mean": 1.1667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8857,
          "delta_mean": 0.4571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 1.5,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.3926,
        "decode_tps_c4": 8.2145,
        "ttft_s": 1.3651,
        "prefill_tps_4k": 62.0325,
        "prefill_tps_12k": 59.6169,
        "load_seconds": 243.2704,
        "size_gb": 52.3497,
        "vram_gb": 52.3497,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "zai-org-glm-4-5-air",
      "display_name": "GLM-4.5-Air",
      "family": "glm",
      "org": "zai-org",
      "params_b": 110.47,
      "arch": "moe",
      "hf_downloads": 454649,
      "hf_rank": 336,
      "active_params_b": 12.0,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/unsloth/GLM-4.5-Air-GGUF:Q4_K_M",
      "size_bytes": 72975748256,
      "model_version_id": 69,
      "run_id": "r20261015-bv1-5ac997",
      "battery_version": "v1",
      "tested_at": "2026-10-15T13:45:35.381Z",
      "overall": 92.7568,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.4133,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.1664,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0569,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 3.5507,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 2,
      "concurrency_max": 2,
      "concurrency_text": "2 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.40.4, flash attention, 16k context; 2 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.66,
      "ci_high": 95.61,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 94.3441,
      "n_truncated": 27,
      "n_extended": 24,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 30,
      "class_rank": 6,
      "rank_lo": 6,
      "rank_hi": 68,
      "class_rank_lo": 2,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 82.8571,
        "math": 96.0,
        "coding": 87.4541,
        "instruction": 94.3333,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 95.0,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8745,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9,
          "delta_mean": 1.1333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9433,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1333,
          "delta_mean": 0.7,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2571,
          "delta_mean": 0.7429,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5417,
          "delta_mean": 1.5833,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 5.9864,
        "decode_tps_c4": 7.5927,
        "ttft_s": 3.1499,
        "prefill_tps_4k": 53.4418,
        "prefill_tps_12k": 28.8813,
        "load_seconds": 110.5245,
        "size_gb": 76.7538,
        "vram_gb": 76.7538,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-30b-a3b",
      "display_name": "Qwen3-30B-A3B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": 1446040,
      "hf_rank": 106,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:30b-a3b-q4_K_M",
      "size_bytes": 18622549504,
      "model_version_id": 45,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.7502,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.784,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.2643,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9245,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 3.4846,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.38,
      "ci_high": 95.57,
      "hard_score": 65.816,
      "hard_n": 38,
      "ext_overall": 93.3851,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 31,
      "class_rank": 21,
      "rank_lo": 8,
      "rank_hi": 65,
      "class_rank_lo": 6,
      "class_rank_hi": 37,
      "categories": {
        "reasoning": 88.5714,
        "math": 84.0,
        "coding": 86.5689,
        "instruction": 92.2778,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8657,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5667,
          "delta_mean": 1.0667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9228,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 0.9667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1714,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 0.625,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 19.6786,
        "decode_tps_c4": 25.4957,
        "ttft_s": 0.5183,
        "prefill_tps_4k": 379.7907,
        "prefill_tps_12k": 301.0756,
        "load_seconds": 101.3695,
        "size_gb": 31.7459,
        "vram_gb": 31.7459,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-3-12b-it",
      "display_name": "gemma-3-12b-it",
      "family": "gemma",
      "org": "google",
      "params_b": 12.19,
      "arch": "dense",
      "hf_downloads": 452772,
      "hf_rank": 338,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3:12b-it-q4_K_M",
      "size_bytes": 8149180896,
      "model_version_id": 6,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.742,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 7.431,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0826,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9864,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 2.5186,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.65,
      "ci_high": 95.37,
      "hard_score": 66.5313,
      "hard_n": 38,
      "ext_overall": 93.0594,
      "n_truncated": 23,
      "n_extended": 20,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 32,
      "class_rank": 2,
      "rank_lo": 8,
      "rank_hi": 66,
      "class_rank_lo": 2,
      "class_rank_hi": 9,
      "categories": {
        "reasoning": 88.5714,
        "math": 96.0,
        "coding": 81.8285,
        "instruction": 88.2778,
        "knowledge": 96.6667,
        "long_context": 83.3333,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8183,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7667,
          "delta_mean": 0.7,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8828,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9333,
          "delta_mean": 1.2,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1714,
          "delta_mean": 0.6571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 1.25,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.1155,
        "decode_tps_c4": 9.4774,
        "ttft_s": 0.843,
        "prefill_tps_4k": 139.0193,
        "prefill_tps_12k": 138.4609,
        "load_seconds": 115.3567,
        "size_gb": 13.0607,
        "vram_gb": 13.0607,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistralai-ministral-3-14b-instruct-2512",
      "display_name": "Ministral 3 14B Instruct (2512)",
      "family": "mistral",
      "org": "mistralai",
      "params_b": 14.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "ministral-3:14b-instruct-2512-q4_K_M",
      "size_bytes": 9082522240,
      "model_version_id": 26,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.7343,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.4357,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.9957,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0814,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 2.324,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.97,
      "ci_high": 95.17,
      "hard_score": 62.3918,
      "hard_n": 38,
      "ext_overall": 94.0042,
      "n_truncated": 16,
      "n_extended": 16,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        31,
        34,
        35,
        36,
        38,
        39,
        40,
        41,
        42,
        43,
        48,
        52,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 33,
      "class_rank": 3,
      "rank_lo": 9,
      "rank_hi": 61,
      "class_rank_lo": 2,
      "class_rank_hi": 8,
      "categories": {
        "reasoning": 82.8571,
        "math": 88.0,
        "coding": 87.5296,
        "instruction": 81.2222,
        "knowledge": 96.6667,
        "long_context": 100.0,
        "multilingual": 98.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8753,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5333,
          "delta_mean": 0.7333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8122,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8667,
          "delta_mean": 1.5667,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8286,
          "delta_mean": 0.8571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.375,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 5.6275,
        "decode_tps_c4": 15.0083,
        "ttft_s": 0.7308,
        "prefill_tps_4k": 177.7442,
        "prefill_tps_12k": 134.7123,
        "load_seconds": 173.0056,
        "size_gb": 29.9808,
        "vram_gb": 29.9808,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwq-32b",
      "display_name": "QwQ 32B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 32.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwq:32b-q4_K_M",
      "size_bytes": 19851336480,
      "model_version_id": 239,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 92.6424,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.4152,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0067,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8376,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 1.1018,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.82,
      "ci_high": 95.35,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 97.6784,
      "n_truncated": 17,
      "n_extended": 14,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        228,
        229,
        230,
        231,
        232,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 34,
      "class_rank": 22,
      "rank_lo": 6,
      "rank_hi": 65,
      "class_rank_lo": 4,
      "class_rank_hi": 35,
      "categories": {
        "reasoning": 94.2857,
        "math": 88.0,
        "coding": 60.6074,
        "instruction": 95.0556,
        "knowledge": 100.0,
        "long_context": 100.0,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6061,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9506,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9667,
          "delta_mean": 1.1667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6,
          "delta_mean": 0.7714,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1549,
        "decode_tps_c4": 4.7182,
        "ttft_s": 1.4579,
        "prefill_tps_4k": 72.092,
        "prefill_tps_12k": 60.1915,
        "load_seconds": 271.7463,
        "size_gb": 54.2244,
        "vram_gb": 54.2244,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-1-30b",
      "display_name": "Granite4.1 30B",
      "family": "granite",
      "org": null,
      "params_b": 30.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.1:30b-q4_K_M",
      "size_bytes": 17490000000,
      "model_version_id": 207,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 92.3162,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.8376,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4383,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8607,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 1.1097,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.29,
      "ci_high": 95.26,
      "hard_score": 62.1804,
      "hard_n": 35,
      "ext_overall": 92.6336,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 35,
      "class_rank": 23,
      "rank_lo": 6,
      "rank_hi": 66,
      "class_rank_lo": 4,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 82.8571,
        "math": 96.0,
        "coding": 85.9883,
        "instruction": 93.7778,
        "knowledge": 100.0,
        "long_context": 72.2222,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8599,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9378,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.8333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 0.6286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5,
          "delta_mean": 0.9167,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.3861,
        "decode_tps_c4": 5.0011,
        "ttft_s": 1.4861,
        "prefill_tps_4k": 62.4365,
        "prefill_tps_12k": 59.1875,
        "load_seconds": 242.0584,
        "size_gb": 52.3497,
        "vram_gb": 52.3497,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-1-8b",
      "display_name": "Granite4.1 8B",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.1:8b-q4_K_M",
      "size_bytes": 5300000000,
      "model_version_id": 122,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 92.1412,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.6017,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3221,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9405,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 3.3107,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 88.96,
      "ci_high": 95.0,
      "hard_score": 76.7792,
      "hard_n": 71,
      "ext_overall": 92.1412,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 36,
      "class_rank": 4,
      "rank_lo": 6,
      "rank_hi": 73,
      "class_rank_lo": 1,
      "class_rank_hi": 14,
      "categories": {
        "reasoning": 91.4286,
        "math": 88.0,
        "coding": 76.4532,
        "instruction": 93.1111,
        "knowledge": 96.6667,
        "long_context": 94.4444,
        "multilingual": 93.3333,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7645,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9311,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.9,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6857,
          "delta_mean": 0.4857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4167,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.436,
        "decode_tps_c4": 15.9417,
        "ttft_s": 0.5004,
        "prefill_tps_4k": 224.8955,
        "prefill_tps_12k": 183.7273,
        "load_seconds": 189.687,
        "size_gb": 27.1039,
        "vram_gb": 27.1039,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistral-small-24b-instruct-2501",
      "display_name": "Mistral Small 24B Instruct 2501",
      "family": "mistral",
      "org": null,
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral-small:24b-instruct-2501-q4_K_M",
      "size_bytes": 14334000000,
      "model_version_id": 219,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 92.138,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.6371,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.066,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8607,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 1.3755,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 89.3,
      "ci_high": 94.69,
      "hard_score": 53.8471,
      "hard_n": 35,
      "ext_overall": 92.138,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        203,
        204,
        206,
        209,
        210,
        214,
        215,
        216,
        218,
        221,
        225,
        226,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 37,
      "class_rank": 24,
      "rank_lo": 6,
      "rank_hi": 65,
      "class_rank_lo": 4,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 85.7143,
        "math": 84.0,
        "coding": 82.7502,
        "instruction": 84.8333,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 97.5,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8275,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6333,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8483,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9333,
          "delta_mean": 0.9333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.975,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9714,
          "delta_mean": 0.6286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.2181,
        "decode_tps_c4": 10.5787,
        "ttft_s": 1.032,
        "prefill_tps_4k": 98.4382,
        "prefill_tps_12k": 87.8892,
        "load_seconds": 204.3266,
        "size_gb": 36.4636,
        "vram_gb": 36.4636,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistralai-devstral-small-2-24b-instruct-2512",
      "display_name": "Devstral Small 2 24B Instruct (2512)",
      "family": "mistral",
      "org": "mistralai",
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "devstral-small-2:24b-instruct-2512-q4_K_M",
      "size_bytes": 15177370240,
      "model_version_id": 25,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.0736,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.8129,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.2126,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0198,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 1.3216,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 89.31,
      "ci_high": 94.71,
      "hard_score": 63.1606,
      "hard_n": 38,
      "ext_overall": 93.2799,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        231,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 38,
      "class_rank": 25,
      "rank_lo": 9,
      "rank_hi": 67,
      "class_rank_lo": 7,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 85.7143,
        "math": 76.0,
        "coding": 83.3924,
        "instruction": 89.1111,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8339,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6333,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8911,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1667,
          "delta_mean": 1.1667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9714,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 1.125,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1363,
        "decode_tps_c4": 8.8856,
        "ttft_s": 1.5395,
        "prefill_tps_4k": 79.9954,
        "prefill_tps_12k": 64.8535,
        "load_seconds": 177.645,
        "size_gb": 36.4783,
        "vram_gb": 36.4783,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-5-9b",
      "display_name": "Qwen3.5-9B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 9.65,
      "arch": "dense",
      "hf_downloads": 8827840,
      "hf_rank": 8,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.5:9b-q4_K_M",
      "size_bytes": 5629109120,
      "model_version_id": 53,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 92.039,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.7495,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.4607,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1369,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 8.3502,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 88.99,
      "ci_high": 95.12,
      "hard_score": 59.8885,
      "hard_n": 38,
      "ext_overall": 95.2418,
      "n_truncated": 17,
      "n_extended": 17,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 39,
      "class_rank": 4,
      "rank_lo": 11,
      "rank_hi": 69,
      "class_rank_lo": 2,
      "class_rank_hi": 10,
      "categories": {
        "reasoning": 85.7143,
        "math": 84.0,
        "coding": 77.3588,
        "instruction": 94.0556,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 98.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7736,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9406,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7667,
          "delta_mean": 1.0333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2286,
          "delta_mean": 0.5429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5833,
          "delta_mean": 1.7083,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.125,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.4176,
        "decode_tps_c4": 8.2368,
        "ttft_s": 0.6868,
        "prefill_tps_4k": 225.4248,
        "prefill_tps_12k": 221.6891,
        "load_seconds": 13.5315,
        "size_gb": 6.1139,
        "vram_gb": 6.1139,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-14b",
      "display_name": "Qwen3-14B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 14.77,
      "arch": "dense",
      "hf_downloads": 3475452,
      "hf_rank": 44,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:14b-q4_K_M",
      "size_bytes": 9276184896,
      "model_version_id": 44,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 91.707,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.7502,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3095,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9083,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 2.0866,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 88.48,
      "ci_high": 94.73,
      "hard_score": 52.9304,
      "hard_n": 38,
      "ext_overall": 92.3419,
      "n_truncated": 10,
      "n_extended": 10,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        36,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 40,
      "class_rank": 5,
      "rank_lo": 11,
      "rank_hi": 71,
      "class_rank_lo": 2,
      "class_rank_hi": 10,
      "categories": {
        "reasoning": 91.4286,
        "math": 88.0,
        "coding": 73.4897,
        "instruction": 93.5556,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 90.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7349,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8,
          "delta_mean": 0.9333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9356,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0667,
          "delta_mean": 0.6333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5143,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1667,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.9961,
        "decode_tps_c4": 8.6657,
        "ttft_s": 0.8052,
        "prefill_tps_4k": 153.4525,
        "prefill_tps_12k": 112.9015,
        "load_seconds": 171.4608,
        "size_gb": 30.9283,
        "vram_gb": 30.9283,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-8b",
      "display_name": "Qwen3-8B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 8.19,
      "arch": "dense",
      "hf_downloads": 10080256,
      "hf_rank": 5,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:8b-q4_K_M",
      "size_bytes": 5225374496,
      "model_version_id": 57,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 91.5462,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.3138,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.7726,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9883,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 3.4534,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 88.37,
      "ci_high": 94.5,
      "hard_score": 55.4539,
      "hard_n": 38,
      "ext_overall": 91.8637,
      "n_truncated": 9,
      "n_extended": 6,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 41,
      "class_rank": 5,
      "rank_lo": 11,
      "rank_hi": 72,
      "class_rank_lo": 2,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 94.2857,
        "math": 88.0,
        "coding": 70.5749,
        "instruction": 87.7222,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7057,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 0.9333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8772,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9333,
          "delta_mean": 1.1667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3714,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7083,
          "delta_mean": 0.7917,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.2875,
        "decode_tps_c4": 14.3728,
        "ttft_s": 0.4989,
        "prefill_tps_4k": 249.4424,
        "prefill_tps_12k": 202.9387,
        "load_seconds": 148.5704,
        "size_gb": 24.6665,
        "vram_gb": 24.6665,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-14b-instruct",
      "display_name": "Qwen2.5-14B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 14.77,
      "arch": "dense",
      "hf_downloads": 1562286,
      "hf_rank": 97,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:14b-instruct-q4_K_M",
      "size_bytes": 8988110688,
      "model_version_id": 32,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 91.2781,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.6212,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.1121,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0214,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 2.1841,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 88.01,
      "ci_high": 94.22,
      "hard_score": 63.63,
      "hard_n": 38,
      "ext_overall": 91.2781,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 42,
      "class_rank": 6,
      "rank_lo": 11,
      "rank_hi": 72,
      "class_rank_lo": 2,
      "class_rank_hi": 11,
      "categories": {
        "reasoning": 88.5714,
        "math": 84.0,
        "coding": 76.4869,
        "instruction": 91.3333,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 91.6667,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7649,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9133,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4,
          "delta_mean": 1.1667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0857,
          "delta_mean": 0.8571,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5,
          "delta_mean": 1.2917,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 21.67,
        "decode_tps_c4": 59.0027,
        "ttft_s": 0.1831,
        "prefill_tps_4k": 555.2244,
        "prefill_tps_12k": 415.7516,
        "load_seconds": 141.0471,
        "size_gb": 34.8467,
        "vram_gb": 34.8467,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "laguna-xs-2-1",
      "display_name": "Laguna XS 2.1",
      "family": "other",
      "org": null,
      "params_b": null,
      "arch": null,
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "laguna-xs-2.1:q4_K_M",
      "size_bytes": 20274000000,
      "model_version_id": 213,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 91.1849,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.9569,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.6205,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9412,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 4.0226,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": null,
      "size_class_label": null,
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 87.65,
      "ci_high": 94.37,
      "hard_score": 74.3924,
      "hard_n": 35,
      "ext_overall": 91.1849,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 43,
      "class_rank": null,
      "rank_lo": 6,
      "rank_hi": 73,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": 91.4286,
        "math": 88.0,
        "coding": 88.5134,
        "instruction": 93.2778,
        "knowledge": 96.6667,
        "long_context": 94.4444,
        "multilingual": 98.3333,
        "tool_use": 70.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8851,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 0.8333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9328,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 0.9,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.6857,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.625,
          "delta_mean": 1.2083,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 23.1114,
        "decode_tps_c4": 40.0182,
        "ttft_s": 0.642,
        "prefill_tps_4k": 323.363,
        "prefill_tps_12k": 288.9103,
        "load_seconds": 68.3955,
        "size_gb": 23.1855,
        "vram_gb": 23.1855,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-coder-32b-instruct",
      "display_name": "Qwen2.5-Coder-32B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 32.76,
      "arch": "dense",
      "hf_downloads": 863297,
      "hf_rank": 195,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:32b-instruct-q4_K_M",
      "size_bytes": 19851336384,
      "model_version_id": 37,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 90.5357,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.6569,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.355,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8829,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 0.9103,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 87.25,
      "ci_high": 93.76,
      "hard_score": 64.9031,
      "hard_n": 38,
      "ext_overall": 90.5357,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 44,
      "class_rank": 26,
      "rank_lo": 12,
      "rank_hi": 74,
      "class_rank_lo": 8,
      "class_rank_hi": 40,
      "categories": {
        "reasoning": 80.0,
        "math": 64.0,
        "coding": 86.5439,
        "instruction": 94.8333,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 95.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8654,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1333,
          "delta_mean": 0.6667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9483,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2,
          "delta_mean": 0.8333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4,
          "delta_mean": 0.9143,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2917,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.3009,
        "decode_tps_c4": 7.6424,
        "ttft_s": 1.6116,
        "prefill_tps_4k": 51.762,
        "prefill_tps_12k": 45.8706,
        "load_seconds": 244.5076,
        "size_gb": 54.2034,
        "vram_gb": 54.2034,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "bartowski-nanbeige-nanbeige4-3-33b",
      "display_name": "Nanbeige_Nanbeige4.3-33B",
      "family": "other",
      "org": null,
      "params_b": 33.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/bartowski/Nanbeige_Nanbeige4.3-33B-GGUF:Q4_K_M",
      "size_bytes": 20000000000,
      "model_version_id": 197,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 90.3666,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.5914,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.1876,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.921,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 1.0683,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.83,
      "ci_high": 93.71,
      "hard_score": 66.1958,
      "hard_n": 35,
      "ext_overall": 90.8913,
      "n_truncated": 20,
      "n_extended": 19,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 45,
      "class_rank": 27,
      "rank_lo": 9,
      "rank_hi": 76,
      "class_rank_lo": 5,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 88.5714,
        "math": 80.0,
        "coding": 90.1173,
        "instruction": 80.7222,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 90.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9012,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7,
          "delta_mean": 0.9333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8072,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6333,
          "delta_mean": 0.9643,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4857,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2857,
          "delta_mean": 0.8571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.0976,
        "decode_tps_c4": 7.3638,
        "ttft_s": 1.5413,
        "prefill_tps_4k": 70.8218,
        "prefill_tps_12k": 58.1961,
        "load_seconds": 200.9046,
        "size_gb": 48.8722,
        "vram_gb": 48.8722,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nemotron-3-5-bolt-e4b",
      "display_name": "Nemotron 3.5 Bolt E4B",
      "family": "nemotron",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 4.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron-3.5-bolt:e4b-q4_K_M",
      "size_bytes": 7200000000,
      "model_version_id": 223,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 90.2903,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.8229,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.56,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.6952,
      "agreement": "high",
      "disputed": 1,
      "tokens_per_s": 3.1475,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.28,
      "ci_high": 93.88,
      "hard_score": 57.0286,
      "hard_n": 35,
      "ext_overall": 93.7295,
      "n_truncated": 11,
      "n_extended": 11,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 46,
      "class_rank": 6,
      "rank_lo": 8,
      "rank_hi": 82,
      "class_rank_lo": 1,
      "class_rank_hi": 22,
      "categories": {
        "reasoning": 82.8571,
        "math": 96.0,
        "coding": 73.2,
        "instruction": 100.0,
        "knowledge": 100.0,
        "long_context": 72.2222,
        "multilingual": 98.3333,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.732,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8667,
          "delta_mean": 0.4,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 0.5429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8333,
          "delta_mean": 0.5417,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.7681,
        "decode_tps_c4": 12.995,
        "ttft_s": 0.3939,
        "prefill_tps_4k": 383.7537,
        "prefill_tps_12k": 364.5984,
        "load_seconds": 41.6602,
        "size_gb": 5.7356,
        "vram_gb": 5.7356,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-4b-instruct-2507",
      "display_name": "Qwen3-4B-Instruct-2507",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 4.02,
      "arch": "dense",
      "hf_downloads": 3648462,
      "hf_rank": 42,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:4b-instruct-2507-q4_K_M",
      "size_bytes": 2497280480,
      "model_version_id": 49,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 90.2603,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.4067,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.1169,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0698,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 14.7613,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 87.17,
      "ci_high": 93.34,
      "hard_score": 57.238,
      "hard_n": 38,
      "ext_overall": 92.5919,
      "n_truncated": 21,
      "n_extended": 21,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        23,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 47,
      "class_rank": 7,
      "rank_lo": 13,
      "rank_hi": 73,
      "class_rank_lo": 2,
      "class_rank_hi": 13,
      "categories": {
        "reasoning": 82.8571,
        "math": 80.0,
        "coding": 69.7079,
        "instruction": 94.2222,
        "knowledge": 96.6667,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6971,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7333,
          "delta_mean": 0.9333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9422,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2,
          "delta_mean": 0.6667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9429,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9583,
          "delta_mean": 1.4167,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.875,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.0756,
        "decode_tps_c4": 43.182,
        "ttft_s": 0.2618,
        "prefill_tps_4k": 412.3869,
        "prefill_tps_12k": 308.3372,
        "load_seconds": 85.7297,
        "size_gb": 12.5367,
        "vram_gb": 12.5367,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "cogito-14b-v1-preview-qwen",
      "display_name": "Cogito 14B V1 Preview Qwen",
      "family": "qwen",
      "org": null,
      "params_b": 14.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2.5",
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "cogito:14b-v1-preview-qwen-q4_K_M",
      "size_bytes": 9000000000,
      "model_version_id": 78,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 90.2542,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.614,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0074,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.921,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 1.7691,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.78,
      "ci_high": 93.38,
      "hard_score": 72.8179,
      "hard_n": 71,
      "ext_overall": 90.2542,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 48,
      "class_rank": 7,
      "rank_lo": 11,
      "rank_hi": 79,
      "class_rank_lo": 2,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 82.8571,
        "math": 84.0,
        "coding": 78.8199,
        "instruction": 89.3889,
        "knowledge": 96.6667,
        "long_context": 88.8889,
        "multilingual": 91.6667,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7882,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5333,
          "delta_mean": 0.7667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8939,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0,
          "delta_mean": 1.4,
          "disputed": 9,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6286,
          "delta_mean": 0.7714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.125,
          "delta_mean": 0.9167,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.6269,
        "decode_tps_c4": 10.208,
        "ttft_s": 0.7463,
        "prefill_tps_4k": 140.3235,
        "prefill_tps_12k": 105.7577,
        "load_seconds": 200.328,
        "size_gb": 34.887,
        "vram_gb": 34.887,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "medgemma-27b-it",
      "display_name": "Medgemma 27B IT",
      "family": "gemma",
      "org": null,
      "params_b": 27.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": "Gemma 3",
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "medgemma:27b-it-q4_K_M",
      "size_bytes": 17397000000,
      "model_version_id": 217,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 90.2453,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.1886,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.009,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0462,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 1.1733,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "medical",
      "specialised_label": "medical",
      "withheld": null,
      "ci_low": 86.71,
      "ci_high": 93.55,
      "hard_score": 49.1897,
      "hard_n": 35,
      "ext_overall": 92.8309,
      "n_truncated": 25,
      "n_extended": 25,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        206,
        209,
        210,
        215,
        216,
        218,
        221,
        225,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 49,
      "class_rank": 28,
      "rank_lo": 10,
      "rank_hi": 75,
      "class_rank_lo": 6,
      "class_rank_hi": 37,
      "categories": {
        "reasoning": 82.8571,
        "math": 84.0,
        "coding": 76.962,
        "instruction": 89.5,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 95.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7696,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9667,
          "delta_mean": 1.0,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.895,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2,
          "delta_mean": 1.0667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6286,
          "delta_mean": 0.9143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 1.25,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.5467,
        "decode_tps_c4": 5.3201,
        "ttft_s": 1.6361,
        "prefill_tps_4k": 87.7612,
        "prefill_tps_12k": 89.0226,
        "load_seconds": 126.635,
        "size_gb": 22.0907,
        "vram_gb": 22.0907,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistral-small3-1-24b-instruct-2503",
      "display_name": "Mistral Small3.1 24B Instruct 2503",
      "family": "mistral",
      "org": null,
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral-small3.1:24b-instruct-2503-q4_K_M",
      "size_bytes": 15487000000,
      "model_version_id": 220,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 90.1023,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 7.4145,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.975,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0505,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 1.3137,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.85,
      "ci_high": 93.06,
      "hard_score": 49.569,
      "hard_n": 35,
      "ext_overall": 90.1023,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 50,
      "class_rank": 29,
      "rank_lo": 12,
      "rank_hi": 77,
      "class_rank_lo": 8,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 88.5714,
        "math": 76.0,
        "coding": 73.2934,
        "instruction": 84.1667,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7329,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9667,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7,
          "delta_mean": 1.1,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 0.6857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7083,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.3107,
        "decode_tps_c4": 10.2841,
        "ttft_s": 1.3437,
        "prefill_tps_4k": 78.7303,
        "prefill_tps_12k": 75.4475,
        "load_seconds": 154.515,
        "size_gb": 36.6328,
        "vram_gb": 36.6328,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "ornith-9b",
      "display_name": "Ornith 9B",
      "family": "other",
      "org": null,
      "params_b": 9.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "ornith:9b-q4_K_M",
      "size_bytes": 5600000000,
      "model_version_id": 147,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 89.9782,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.8176,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.3902,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8421,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 10.7161,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.49,
      "ci_high": 93.06,
      "hard_score": 67.9283,
      "hard_n": 71,
      "ext_overall": 91.0575,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 51,
      "class_rank": 8,
      "rank_lo": 12,
      "rank_hi": 79,
      "class_rank_lo": 3,
      "class_rank_hi": 17,
      "categories": {
        "reasoning": 91.4286,
        "math": 84.0,
        "coding": 63.0415,
        "instruction": 91.8889,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 85.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6304,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9189,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 0.7,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7429,
          "delta_mean": 0.4857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.8333,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.7962,
        "decode_tps_c4": 10.5076,
        "ttft_s": 0.797,
        "prefill_tps_4k": 261.6314,
        "prefill_tps_12k": 256.3338,
        "load_seconds": 12.312,
        "size_gb": 5.8873,
        "vram_gb": 5.8873,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-vl-4b-instruct",
      "display_name": "Qwen3-VL-4B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 4.44,
      "arch": "dense",
      "hf_downloads": 3407511,
      "hf_rank": 47,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:4b-instruct-q4_K_M",
      "size_bytes": 3295612928,
      "model_version_id": 61,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 89.8952,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.0733,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.6733,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0362,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 16.9857,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.43,
      "ci_high": 93.08,
      "hard_score": 57.5917,
      "hard_n": 38,
      "ext_overall": 90.9371,
      "n_truncated": 30,
      "n_extended": 30,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        35,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 52,
      "class_rank": 9,
      "rank_lo": 12,
      "rank_hi": 76,
      "class_rank_lo": 2,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 91.4286,
        "math": 92.0,
        "coding": 61.1558,
        "instruction": 96.5556,
        "knowledge": 90.0,
        "long_context": 83.3333,
        "multilingual": 96.6667,
        "tool_use": 100.0,
        "safety": 97.9167,
        "sycophancy": 75.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6116,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0333,
          "delta_mean": 0.8667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2667,
          "delta_mean": 0.7667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4,
          "delta_mean": 0.7143,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9792,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.3333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 1.5,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.3327,
        "decode_tps_c4": 16.775,
        "ttft_s": 0.2484,
        "prefill_tps_4k": 455.209,
        "prefill_tps_12k": 363.7351,
        "load_seconds": 23.8218,
        "size_gb": 5.4601,
        "vram_gb": 5.4601,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-vl-32b-instruct",
      "display_name": "Qwen2.5-VL-32B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 33.45,
      "arch": "dense",
      "hf_downloads": 792676,
      "hf_rank": 204,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5vl:32b-q4_K_M",
      "size_bytes": 21159298208,
      "model_version_id": 40,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 89.7263,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 6.9364,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.636,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 2.3993,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.53,
      "ci_high": 92.78,
      "hard_score": 58.3057,
      "hard_n": 38,
      "ext_overall": 89.7263,
      "n_truncated": 10,
      "n_extended": 10,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        9,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 53,
      "class_rank": 30,
      "rank_lo": 19,
      "rank_hi": 77,
      "class_rank_lo": 13,
      "class_rank_hi": 40,
      "categories": {
        "reasoning": 88.5714,
        "math": 84.0,
        "coding": 86.5761,
        "instruction": 69.5,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 90.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8658,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0333,
          "delta_mean": 0.7667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.695,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8333,
          "delta_mean": 1.4333,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7714,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9167,
          "delta_mean": 1.0,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.3122,
        "decode_tps_c4": 2.6575,
        "ttft_s": 1.6133,
        "prefill_tps_4k": 51.5838,
        "prefill_tps_12k": null,
        "load_seconds": 248.3929,
        "size_gb": 54.9396,
        "vram_gb": 54.9396,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "translategemma-12b-it",
      "display_name": "Translategemma 12B IT",
      "family": "gemma",
      "org": null,
      "params_b": 12.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": "Gemma 3",
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "translategemma:12b-it-q4_K_M",
      "size_bytes": 8100000000,
      "model_version_id": 182,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 89.6385,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.2671,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.9007,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.954,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 2.2455,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": "translation",
      "specialised_label": "translation",
      "withheld": null,
      "ci_low": 86.37,
      "ci_high": 92.86,
      "hard_score": 71.6904,
      "hard_n": 71,
      "ext_overall": 90.2734,
      "n_truncated": 18,
      "n_extended": 18,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 54,
      "class_rank": 8,
      "rank_lo": 13,
      "rank_hi": 80,
      "class_rank_lo": 2,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 82.8571,
        "math": 92.0,
        "coding": 78.9451,
        "instruction": 86.2778,
        "knowledge": 100.0,
        "long_context": 66.6667,
        "multilingual": 100.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7895,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 1.0667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8628,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1,
          "delta_mean": 0.8667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8286,
          "delta_mean": 0.6286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 1.2083,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 7.6397,
        "decode_tps_c4": 11.538,
        "ttft_s": 0.7732,
        "prefill_tps_4k": 176.8125,
        "prefill_tps_12k": 175.8798,
        "load_seconds": 97.297,
        "size_gb": 13.0755,
        "vram_gb": 13.0755,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "olmo-3-1-32b-instruct",
      "display_name": "Olmo 3.1 32B Instruct",
      "family": "olmo",
      "org": null,
      "params_b": 32.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "olmo-3.1:32b-instruct-q4_K_M",
      "size_bytes": 19482000000,
      "model_version_id": 227,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 89.5821,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.7645,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.5507,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9243,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 0.953,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 86.03,
      "ci_high": 93.19,
      "hard_score": 54.232,
      "hard_n": 35,
      "ext_overall": 93.0865,
      "n_truncated": 11,
      "n_extended": 11,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        198,
        199,
        200,
        201,
        202,
        204,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 55,
      "class_rank": 31,
      "rank_lo": 10,
      "rank_hi": 82,
      "class_rank_lo": 6,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 82.8571,
        "math": 84.0,
        "coding": 84.7151,
        "instruction": 95.2222,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 90.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8472,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0667,
          "delta_mean": 1.1667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9522,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2333,
          "delta_mean": 0.7667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8286,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.625,
          "delta_mean": 0.6667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1416,
        "decode_tps_c4": 7.3567,
        "ttft_s": 2.6116,
        "prefill_tps_4k": 55.5619,
        "prefill_tps_12k": 53.7029,
        "load_seconds": 128.4439,
        "size_gb": 27.7104,
        "vram_gb": 27.7104,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-4b",
      "display_name": "Qwen3-4B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 4.02,
      "arch": "dense",
      "hf_downloads": 7637056,
      "hf_rank": 13,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:4b-q4_K_M",
      "size_bytes": 2620774592,
      "model_version_id": 48,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 89.1828,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.1557,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.8224,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1243,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 9.7572,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.55,
      "ci_high": 92.43,
      "hard_score": 58.0533,
      "hard_n": 38,
      "ext_overall": 89.8177,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 56,
      "class_rank": 10,
      "rank_lo": 18,
      "rank_hi": 81,
      "class_rank_lo": 3,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 68.9627,
        "instruction": 91.0,
        "knowledge": 100.0,
        "long_context": 77.7778,
        "multilingual": 88.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6896,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2333,
          "delta_mean": 1.0667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.91,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2,
          "delta_mean": 1.2333,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4286,
          "delta_mean": 0.5714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 1.75,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 16.9503,
        "decode_tps_c4": 41.3392,
        "ttft_s": 0.2688,
        "prefill_tps_4k": 483.6555,
        "prefill_tps_12k": 342.9295,
        "load_seconds": 66.5486,
        "size_gb": 12.6602,
        "vram_gb": 12.6602,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-ai-deepseek-r1-distill-qwen-32b",
      "display_name": "DeepSeek-R1-Distill-Qwen-32B",
      "family": "deepseek",
      "org": "deepseek-ai",
      "params_b": 32.76,
      "arch": "dense",
      "hf_downloads": 430537,
      "hf_rank": 355,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:32b-qwen-distill-q4_K_M",
      "size_bytes": 19851335552,
      "model_version_id": 232,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 89.0854,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 6.5521,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.2026,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8929,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 1.0492,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.96,
      "ci_high": 91.93,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 92.2087,
      "n_truncated": 15,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        189,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236,
        238
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 57,
      "class_rank": 32,
      "rank_lo": 19,
      "rank_hi": 85,
      "class_rank_lo": 14,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 97.1429,
        "math": 92.0,
        "coding": 53.5704,
        "instruction": 75.7222,
        "knowledge": 100.0,
        "long_context": 100.0,
        "multilingual": 96.6667,
        "tool_use": 95.0,
        "safety": 91.6667,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5357,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.3333,
          "delta_mean": 0.4667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7572,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3667,
          "delta_mean": 1.5667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7714,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 0.7917,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.095,
        "decode_tps_c4": 4.7697,
        "ttft_s": 1.3583,
        "prefill_tps_4k": 56.4769,
        "prefill_tps_12k": 48.8082,
        "load_seconds": 281.5239,
        "size_gb": 19.8513,
        "vram_gb": 19.8513,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "flint-40b-a2b",
      "display_name": "Flint 40B A2B",
      "family": "other",
      "org": null,
      "params_b": 40.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 2.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "NVFP4",
      "ollama_ref": "flint:40b-a2b",
      "size_bytes": 23000000000,
      "model_version_id": 205,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 88.9758,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.776,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.091,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9367,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 5.1286,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.14,
      "ci_high": 92.57,
      "hard_score": 75.1601,
      "hard_n": 35,
      "ext_overall": 92.4855,
      "n_truncated": 24,
      "n_extended": 24,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 58,
      "class_rank": 7,
      "rank_lo": 14,
      "rank_hi": 90,
      "class_rank_lo": 4,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 71.4286,
        "math": 100.0,
        "coding": 93.3534,
        "instruction": 96.5556,
        "knowledge": 90.0,
        "long_context": 77.7778,
        "multilingual": 91.6667,
        "tool_use": 80.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9335,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9667,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4,
          "delta_mean": 0.7333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 0.7714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2381,
          "delta_mean": 1.0952,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 25.8941,
        "decode_tps_c4": 31.8972,
        "ttft_s": 0.6642,
        "prefill_tps_4k": 342.457,
        "prefill_tps_12k": 322.8328,
        "load_seconds": 60.3351,
        "size_gb": 24.2801,
        "vram_gb": 24.2801,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "ornith-ai-ornith-1-5-9b",
      "display_name": "Ornith-1.5-9B",
      "family": "other",
      "org": "ornith-ai",
      "params_b": 9.65,
      "arch": "dense",
      "hf_downloads": 450460,
      "hf_rank": 341,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/bartowski/Ornith-1.5-9B-GGUF:Q4_K_M",
      "size_bytes": null,
      "model_version_id": 31,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 88.9704,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 0,
      "judge_a_mean": 7.5688,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0014,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9021,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 9.6995,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.44,
      "ci_high": 92.38,
      "hard_score": 52.1485,
      "hard_n": 38,
      "ext_overall": 89.5876,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        30,
        34,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 59,
      "class_rank": 9,
      "rank_lo": 19,
      "rank_hi": 81,
      "class_rank_lo": 3,
      "class_rank_hi": 13,
      "categories": {
        "reasoning": 91.4286,
        "math": 84.0,
        "coding": 67.8603,
        "instruction": 85.2222,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 88.3333,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 100.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6786,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3667,
          "delta_mean": 1.0667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8522,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 1.3,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4571,
          "delta_mean": 0.6857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7083,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.1416,
        "decode_tps_c4": 7.9859,
        "ttft_s": 0.7055,
        "prefill_tps_4k": 219.6858,
        "prefill_tps_12k": 214.817,
        "load_seconds": 11.6635,
        "size_gb": 6.3955,
        "vram_gb": 6.3955,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "devstral-24b-small-2505",
      "display_name": "Devstral 24B Small 2505",
      "family": "mistral",
      "org": null,
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "devstral:24b-small-2505-q4_K_M",
      "size_bytes": 14334000000,
      "model_version_id": 203,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 88.945,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.2981,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.764,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9388,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 1.3307,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 85.58,
      "ci_high": 91.95,
      "hard_score": 40.1198,
      "hard_n": 35,
      "ext_overall": 88.945,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        18,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        218,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 60,
      "class_rank": 33,
      "rank_lo": 23,
      "rank_hi": 86,
      "class_rank_lo": 17,
      "class_rank_hi": 40,
      "categories": {
        "reasoning": 77.1429,
        "math": 76.0,
        "coding": 70.4175,
        "instruction": 84.1667,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 98.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7042,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7333,
          "delta_mean": 0.8,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3667,
          "delta_mean": 1.5,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4286,
          "delta_mean": 0.6857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4167,
          "delta_mean": 0.7083,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.1843,
        "decode_tps_c4": 8.0061,
        "ttft_s": 1.5394,
        "prefill_tps_4k": 121.9437,
        "prefill_tps_12k": 86.047,
        "load_seconds": 165.404,
        "size_gb": 36.4636,
        "vram_gb": 36.4636,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "zai-org-glm-4-7-flash",
      "display_name": "GLM-4.7-Flash",
      "family": "glm",
      "org": "zai-org",
      "params_b": 31.22,
      "arch": "moe",
      "hf_downloads": 1677746,
      "hf_rank": 87,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "glm-4.7-flash:q4_K_M",
      "size_bytes": 19019269280,
      "model_version_id": 64,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 88.9114,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 7.145,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.73,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1693,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 2.5924,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.16,
      "ci_high": 92.47,
      "hard_score": 60.2667,
      "hard_n": 38,
      "ext_overall": 90.6257,
      "n_truncated": 12,
      "n_extended": 12,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 61,
      "class_rank": 34,
      "rank_lo": 17,
      "rank_hi": 84,
      "class_rank_lo": 12,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 77.1429,
        "math": 92.0,
        "coding": 81.56,
        "instruction": 86.7222,
        "knowledge": 100.0,
        "long_context": 77.7778,
        "multilingual": 100.0,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8156,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5333,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8672,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9333,
          "delta_mean": 1.0333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 0.9714,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9583,
          "delta_mean": 1.5,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.375,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 14.1416,
        "decode_tps_c4": 19.3114,
        "ttft_s": 0.5803,
        "prefill_tps_4k": 186.6383,
        "prefill_tps_12k": 104.3385,
        "load_seconds": 63.9481,
        "size_gb": 26.0261,
        "vram_gb": 26.0261,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-5-4b",
      "display_name": "Qwen3.5-4B",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 4.66,
      "arch": "dense",
      "hf_downloads": 7968478,
      "hf_rank": 12,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.5:4b-q4_K_M",
      "size_bytes": 2648593216,
      "model_version_id": 52,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 88.7306,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.315,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.9457,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1155,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 14.974,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 84.93,
      "ci_high": 92.35,
      "hard_score": 55.78,
      "hard_n": 38,
      "ext_overall": 91.362,
      "n_truncated": 16,
      "n_extended": 16,
      "n_gated": 0,
      "beats": [
        1,
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 62,
      "class_rank": 11,
      "rank_lo": 19,
      "rank_hi": 84,
      "class_rank_lo": 3,
      "class_rank_hi": 20,
      "categories": {
        "reasoning": 85.7143,
        "math": 88.0,
        "coding": 71.4721,
        "instruction": 90.8889,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 98.3333,
        "tool_use": 85.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7147,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8333,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9089,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1667,
          "delta_mean": 1.3333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2286,
          "delta_mean": 0.4857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.625,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 15.0613,
        "decode_tps_c4": 14.6499,
        "ttft_s": 0.4221,
        "prefill_tps_4k": 359.9639,
        "prefill_tps_12k": 347.9449,
        "load_seconds": 11.0953,
        "size_gb": 3.6291,
        "vram_gb": 3.6291,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "ministral-3-8b-instruct-2512",
      "display_name": "Ministral 3 8B Instruct 2512",
      "family": "mistral",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "ministral-3:8b-instruct-2512-q4_K_M",
      "size_bytes": 6000000000,
      "model_version_id": 139,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 88.7303,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.111,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.6543,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.999,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 3.4288,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 85.23,
      "ci_high": 92.08,
      "hard_score": 68.8468,
      "hard_n": 71,
      "ext_overall": 88.8572,
      "n_truncated": 17,
      "n_extended": 17,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        17,
        18,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        80,
        81,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        89,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        104,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        127,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        138,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 63,
      "class_rank": 12,
      "rank_lo": 19,
      "rank_hi": 86,
      "class_rank_lo": 4,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 82.8571,
        "math": 80.0,
        "coding": 78.4374,
        "instruction": 77.8333,
        "knowledge": 96.6667,
        "long_context": 94.4444,
        "multilingual": 88.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7844,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5333,
          "delta_mean": 0.7333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7783,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6667,
          "delta_mean": 1.4333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5714,
          "delta_mean": 0.8286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 1.125,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.9993,
        "decode_tps_c4": 22.1928,
        "ttft_s": 0.4141,
        "prefill_tps_4k": 270.9169,
        "prefill_tps_12k": 200.5405,
        "load_seconds": 145.25,
        "size_gb": 23.7624,
        "vram_gb": 23.7624,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-coder-14b-instruct",
      "display_name": "Qwen2.5-Coder-14B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 14.77,
      "arch": "dense",
      "hf_downloads": 1622509,
      "hf_rank": 93,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:14b-instruct-q4_K_M",
      "size_bytes": 8988110784,
      "model_version_id": 36,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 87.7494,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 7.2902,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.734,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.011,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 1.8754,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 83.95,
      "ci_high": 91.25,
      "hard_score": 52.2923,
      "hard_n": 38,
      "ext_overall": 87.7494,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        8,
        13,
        14,
        15,
        16,
        17,
        18,
        19,
        20,
        21,
        22,
        24,
        27,
        29,
        34,
        38,
        39,
        41,
        42,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 64,
      "class_rank": 10,
      "rank_lo": 30,
      "rank_hi": 86,
      "class_rank_lo": 6,
      "class_rank_hi": 13,
      "categories": {
        "reasoning": 82.8571,
        "math": 72.0,
        "coding": 82.6655,
        "instruction": 84.1667,
        "knowledge": 93.3333,
        "long_context": 88.8889,
        "multilingual": 85.8333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8267,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3333,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6667,
          "delta_mean": 1.3,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6286,
          "delta_mean": 0.9714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 5.1637,
        "decode_tps_c4": 15.8646,
        "ttft_s": 0.741,
        "prefill_tps_4k": 112.1146,
        "prefill_tps_12k": 86.4847,
        "load_seconds": 203.3631,
        "size_gb": 34.8467,
        "vram_gb": 34.8467,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lfm2-24b",
      "display_name": "Lfm2 24B",
      "family": "lfm",
      "org": null,
      "params_b": 24.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 2.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "lfm2:24b-q4_K_M",
      "size_bytes": 14416000000,
      "model_version_id": 214,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 87.3294,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.4702,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0017,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8176,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 32.7241,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 83.22,
      "ci_high": 91.12,
      "hard_score": 51.9428,
      "hard_n": 35,
      "ext_overall": 87.3294,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        14,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        210,
        215,
        216,
        221,
        228,
        229,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 65,
      "class_rank": 35,
      "rank_lo": 26,
      "rank_hi": 96,
      "class_rank_lo": 19,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 77.1429,
        "math": 96.0,
        "coding": 72.9333,
        "instruction": 96.5556,
        "knowledge": 100.0,
        "long_context": 66.6667,
        "multilingual": 91.6667,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7293,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3,
          "delta_mean": 0.7667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2,
          "delta_mean": 0.5714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.625,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 32.4672,
        "decode_tps_c4": 29.4636,
        "ttft_s": 0.5077,
        "prefill_tps_4k": 583.0121,
        "prefill_tps_12k": 591.1745,
        "load_seconds": 16.1083,
        "size_gb": 14.9609,
        "vram_gb": 14.9609,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nemotron3-33b",
      "display_name": "Nemotron3 33B",
      "family": "nemotron",
      "org": null,
      "params_b": 33.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron3:33b-q4_K_M",
      "size_bytes": 27639000000,
      "model_version_id": 226,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 86.2321,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.5095,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.2002,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8455,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 20.0037,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 82.02,
      "ci_high": 89.99,
      "hard_score": 65.5997,
      "hard_n": 35,
      "ext_overall": 86.2321,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 66,
      "class_rank": 36,
      "rank_lo": 32,
      "rank_hi": 101,
      "class_rank_lo": 23,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 80.0,
        "math": 84.0,
        "coding": 83.6996,
        "instruction": 92.5556,
        "knowledge": 100.0,
        "long_context": 83.3333,
        "multilingual": 55.0,
        "tool_use": 97.5,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.837,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0667,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9256,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4667,
          "delta_mean": 1.1667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.55,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3429,
          "delta_mean": 0.4857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.125,
          "delta_mean": 1.0833,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.975,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 19.9596,
        "decode_tps_c4": 18.8633,
        "ttft_s": 0.829,
        "prefill_tps_4k": 372.0848,
        "prefill_tps_12k": 366.7239,
        "load_seconds": 24.012,
        "size_gb": 24.9648,
        "vram_gb": 24.9648,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-3n-e4b-it",
      "display_name": "Gemma 3n E4B IT",
      "family": "gemma",
      "org": "google",
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 4.0,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3n:e4b-it-q4_K_M",
      "size_bytes": 7547579904,
      "model_version_id": 9,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 86.1364,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.8688,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3936,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2343,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 4.4371,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 82.37,
      "ci_high": 89.57,
      "hard_score": 50.7309,
      "hard_n": 38,
      "ext_overall": 86.7713,
      "n_truncated": 20,
      "n_extended": 20,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        18,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 67,
      "class_rank": 13,
      "rank_lo": 44,
      "rank_hi": 97,
      "class_rank_lo": 7,
      "class_rank_hi": 29,
      "categories": {
        "reasoning": 82.8571,
        "math": 92.0,
        "coding": 65.2591,
        "instruction": 83.4444,
        "knowledge": 96.6667,
        "long_context": 66.6667,
        "multilingual": 93.3333,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6526,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8344,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1667,
          "delta_mean": 1.4667,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5429,
          "delta_mean": 0.7714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5,
          "delta_mean": 1.2083,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.0698,
        "decode_tps_c4": 20.0344,
        "ttft_s": 0.5395,
        "prefill_tps_4k": 282.0702,
        "prefill_tps_12k": 276.8762,
        "load_seconds": 22.3761,
        "size_gb": 4.2095,
        "vram_gb": 4.2095,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-4",
      "display_name": "phi-4",
      "family": "phi",
      "org": "microsoft",
      "params_b": 14.66,
      "arch": "dense",
      "hf_downloads": 426836,
      "hf_rank": 357,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi4:14b-q4_K_M",
      "size_bytes": 9053114464,
      "model_version_id": 23,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 85.9357,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 7.5226,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.9229,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0107,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 2.1761,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 81.87,
      "ci_high": 89.76,
      "hard_score": 64.2051,
      "hard_n": 38,
      "ext_overall": 86.3061,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 68,
      "class_rank": 11,
      "rank_lo": 39,
      "rank_hi": 99,
      "class_rank_lo": 7,
      "class_rank_hi": 14,
      "categories": {
        "reasoning": 91.4286,
        "math": 84.0,
        "coding": 86.9932,
        "instruction": 73.2222,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 93.3333,
        "tool_use": 70.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8699,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7322,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3,
          "delta_mean": 1.4333,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3143,
          "delta_mean": 0.6286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9167,
          "delta_mean": 1.0,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.9913,
        "decode_tps_c4": 8.9447,
        "ttft_s": 0.7453,
        "prefill_tps_4k": 135.446,
        "prefill_tps_12k": 122.4466,
        "load_seconds": 210.5625,
        "size_gb": 36.1955,
        "vram_gb": 36.1955,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nemotron-cascade-2-30b-a3b",
      "display_name": "Nemotron Cascade 2 30B A3B",
      "family": "nemotron",
      "org": null,
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron-cascade-2:30b-a3b-q4_K_M",
      "size_bytes": 24272000000,
      "model_version_id": 225,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 85.5987,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.0752,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.8298,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9569,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 20.2063,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 81.22,
      "ci_high": 89.7,
      "hard_score": 62.7477,
      "hard_n": 35,
      "ext_overall": 85.5164,
      "n_truncated": 17,
      "n_extended": 16,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 69,
      "class_rank": 37,
      "rank_lo": 36,
      "rank_hi": 106,
      "class_rank_lo": 25,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 68.5714,
        "math": 80.0,
        "coding": 77.8723,
        "instruction": 97.8333,
        "knowledge": 90.0,
        "long_context": 77.7778,
        "multilingual": 78.3333,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7787,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4667,
          "delta_mean": 0.9333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9783,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4667,
          "delta_mean": 0.7667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2571,
          "delta_mean": 0.5429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5833,
          "delta_mean": 1.7917,
          "disputed": 5,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.465,
        "decode_tps_c4": 17.0905,
        "ttft_s": 0.816,
        "prefill_tps_4k": 372.0278,
        "prefill_tps_12k": 367.8146,
        "load_seconds": 22.3707,
        "size_gb": 24.4388,
        "vram_gb": 24.4388,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma2-27b-instruct",
      "display_name": "Gemma2 27B Instruct",
      "family": "gemma",
      "org": null,
      "params_b": 27.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma2:27b-instruct-q4_K_M",
      "size_bytes": 16645000000,
      "model_version_id": 206,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 85.5807,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.2355,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.8186,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9236,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 0.9912,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 81.38,
      "ci_high": 89.41,
      "hard_score": 48.9364,
      "hard_n": 35,
      "ext_overall": 85.5807,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 1,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 70,
      "class_rank": 38,
      "rank_lo": 39,
      "rank_hi": 103,
      "class_rank_lo": 26,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 65.7143,
        "math": 68.0,
        "coding": 67.4561,
        "instruction": 98.2222,
        "knowledge": 96.6667,
        "long_context": 83.3333,
        "multilingual": 95.0,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6746,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2667,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9822,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5667,
          "delta_mean": 0.7667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3429,
          "delta_mean": 0.9429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4167,
          "delta_mean": 0.6667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.375,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.6926,
        "decode_tps_c4": 5.5196,
        "ttft_s": 1.5213,
        "prefill_tps_4k": 67.0124,
        "prefill_tps_12k": 67.9504,
        "load_seconds": 157.6047,
        "size_gb": 27.0161,
        "vram_gb": 27.0161,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-7b-instruct",
      "display_name": "Qwen2.5-7B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 7.62,
      "arch": "dense",
      "hf_downloads": 8272384,
      "hf_rank": 11,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:7b-instruct-q4_K_M",
      "size_bytes": 4683073952,
      "model_version_id": 35,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 85.5675,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.7993,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.2983,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9376,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 6.0586,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 81.38,
      "ci_high": 89.27,
      "hard_score": 44.7327,
      "hard_n": 38,
      "ext_overall": 85.5675,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        15,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        192,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 71,
      "class_rank": 14,
      "rank_lo": 45,
      "rank_hi": 100,
      "class_rank_lo": 7,
      "class_rank_hi": 31,
      "categories": {
        "reasoning": 77.1429,
        "math": 76.0,
        "coding": 63.2422,
        "instruction": 88.4444,
        "knowledge": 100.0,
        "long_context": 77.7778,
        "multilingual": 91.6667,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6324,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8844,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9,
          "delta_mean": 1.1333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7083,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.625,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.1198,
        "decode_tps_c4": 28.2939,
        "ttft_s": 0.3928,
        "prefill_tps_4k": 306.2546,
        "prefill_tps_12k": 278.8864,
        "load_seconds": 34.1411,
        "size_gb": 8.7495,
        "vram_gb": 8.7495,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-ai-deepseek-r1-distill-qwen-14b",
      "display_name": "DeepSeek-R1-Distill-Qwen-14B",
      "family": "deepseek",
      "org": "deepseek-ai",
      "params_b": 14.77,
      "arch": "dense",
      "hf_downloads": 335528,
      "hf_rank": 450,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:14b-qwen-distill-q4_K_M",
      "size_bytes": 8988109952,
      "model_version_id": 231,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 85.2897,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 6.3738,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9457,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.989,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 2.128,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 81.19,
      "ci_high": 89.2,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 88.8904,
      "n_truncated": 15,
      "n_extended": 15,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 72,
      "class_rank": 12,
      "rank_lo": 38,
      "rank_hi": 106,
      "class_rank_lo": 4,
      "class_rank_hi": 16,
      "categories": {
        "reasoning": 91.4286,
        "math": 88.0,
        "coding": 52.9562,
        "instruction": 80.7778,
        "knowledge": 100.0,
        "long_context": 77.7778,
        "multilingual": 89.1667,
        "tool_use": 100.0,
        "safety": 87.5,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5296,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2667,
          "delta_mean": 0.5333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8078,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3,
          "delta_mean": 1.3,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8286,
          "delta_mean": 1.0286,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4583,
          "delta_mean": 1.0833,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 7.1022,
        "decode_tps_c4": 9.9981,
        "ttft_s": 0.6174,
        "prefill_tps_4k": 146.8286,
        "prefill_tps_12k": 104.4517,
        "load_seconds": 173.3613,
        "size_gb": 8.9881,
        "vram_gb": 8.9881,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "exaone3-5-32b-instruct",
      "display_name": "Exaone3.5 32B Instruct",
      "family": "exaone",
      "org": null,
      "params_b": 32.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "exaone3.5:32b-instruct-q4_K_M",
      "size_bytes": 19344000000,
      "model_version_id": 204,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 85.0293,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 7.3505,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.9307,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.906,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 1.0161,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 80.94,
      "ci_high": 88.82,
      "hard_score": 36.7315,
      "hard_n": 35,
      "ext_overall": 85.0293,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 73,
      "class_rank": 39,
      "rank_lo": 46,
      "rank_hi": 106,
      "class_rank_lo": 30,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 77.1429,
        "math": 76.0,
        "coding": 71.51,
        "instruction": 89.7778,
        "knowledge": 96.6667,
        "long_context": 66.6667,
        "multilingual": 87.5,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7151,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7333,
          "delta_mean": 0.8667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8978,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3667,
          "delta_mean": 0.9333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4286,
          "delta_mean": 0.7714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.625,
          "delta_mean": 0.7083,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1191,
        "decode_tps_c4": 7.6686,
        "ttft_s": 1.3083,
        "prefill_tps_4k": 66.3634,
        "prefill_tps_12k": 56.7781,
        "load_seconds": 256.6016,
        "size_gb": 53.8584,
        "vram_gb": 53.8584,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "bartowski-spark-x2-6-4b",
      "display_name": "Spark-X2.6-4B",
      "family": "other",
      "org": null,
      "params_b": 4.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/bartowski/Spark-X2.6-4B-GGUF:Q4_K_M",
      "size_bytes": 2500000000,
      "model_version_id": 198,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 85.0137,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.7637,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.5864,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9324,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 5.0254,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 80.91,
      "ci_high": 89.1,
      "hard_score": 41.7793,
      "hard_n": 35,
      "ext_overall": 89.6054,
      "n_truncated": 29,
      "n_extended": 29,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 74,
      "class_rank": 1,
      "rank_lo": 43,
      "rank_hi": 106,
      "class_rank_lo": 1,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 85.7143,
        "math": 84.0,
        "coding": 65.5758,
        "instruction": 78.7222,
        "knowledge": 90.0,
        "long_context": 77.7778,
        "multilingual": 98.3333,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6558,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7872,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5172,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 0.8621,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.2571,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.5833,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 0.75,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 19.7465,
        "decode_tps_c4": 38.714,
        "ttft_s": 0.3802,
        "prefill_tps_4k": 468.9994,
        "prefill_tps_12k": 438.4377,
        "load_seconds": 58.1939,
        "size_gb": 5.3148,
        "vram_gb": 5.3148,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistral-small-22b-instruct-2409",
      "display_name": "Mistral Small 22B Instruct 2409",
      "family": "mistral",
      "org": null,
      "params_b": 22.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral-small:22b-instruct-2409-q4_K_M",
      "size_bytes": 13341000000,
      "model_version_id": 218,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 84.6347,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.9238,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.2926,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1312,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 1.4163,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 80.66,
      "ci_high": 88.5,
      "hard_score": 36.7692,
      "hard_n": 35,
      "ext_overall": 84.6347,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        196,
        199,
        200,
        201,
        202,
        209,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 75,
      "class_rank": 40,
      "rank_lo": 47,
      "rank_hi": 108,
      "class_rank_lo": 31,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 74.2857,
        "math": 72.0,
        "coding": 70.093,
        "instruction": 78.9444,
        "knowledge": 96.6667,
        "long_context": 88.8889,
        "multilingual": 94.1667,
        "tool_use": 95.0,
        "safety": 91.6667,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7009,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2667,
          "delta_mean": 1.1,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7894,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8667,
          "delta_mean": 1.5,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3714,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 1.0417,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.5,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.2313,
        "decode_tps_c4": 6.1178,
        "ttft_s": 1.1861,
        "prefill_tps_4k": 76.2456,
        "prefill_tps_12k": 64.834,
        "load_seconds": 244.6143,
        "size_gb": 43.9631,
        "vram_gb": 43.9631,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-coder-7b-instruct",
      "display_name": "Qwen2.5-Coder-7B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 7.62,
      "arch": "dense",
      "hf_downloads": 2398003,
      "hf_rank": 62,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:7b-instruct-q4_K_M",
      "size_bytes": 4683074048,
      "model_version_id": 39,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 84.4797,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.5933,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.0795,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1786,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.9176,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 80.57,
      "ci_high": 88.13,
      "hard_score": 42.2492,
      "hard_n": 38,
      "ext_overall": 84.4797,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        2,
        3,
        4,
        5,
        13,
        14,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 76,
      "class_rank": 15,
      "rank_lo": 50,
      "rank_hi": 106,
      "class_rank_lo": 9,
      "class_rank_hi": 32,
      "categories": {
        "reasoning": 82.8571,
        "math": 72.0,
        "coding": 71.7935,
        "instruction": 72.5556,
        "knowledge": 93.3333,
        "long_context": 77.7778,
        "multilingual": 90.0,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 58.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7179,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8,
          "delta_mean": 1.0667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7256,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 1.3333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7143,
          "delta_mean": 0.7429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.625,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.8024,
        "decode_tps_c4": 28.2216,
        "ttft_s": 0.396,
        "prefill_tps_4k": 307.7871,
        "prefill_tps_12k": 279.3404,
        "load_seconds": 62.7367,
        "size_gb": 12.5076,
        "vram_gb": 12.5076,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nvidia-nemotron-3-nano-30b-a3b",
      "display_name": "NVIDIA Nemotron 3 Nano 30B-A3B",
      "family": "nemotron",
      "org": "nvidia",
      "params_b": 30.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 3.0,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron-3-nano:30b-a3b-q4_K_M",
      "size_bytes": 24271924224,
      "model_version_id": 30,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 83.8913,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 7.4314,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0336,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0912,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 18.5296,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 79.5,
      "ci_high": 88.47,
      "hard_score": 62.5237,
      "hard_n": 38,
      "ext_overall": 83.8913,
      "n_truncated": 14,
      "n_extended": 14,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        14,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 77,
      "class_rank": 41,
      "rank_lo": 49,
      "rank_hi": 113,
      "class_rank_lo": 28,
      "class_rank_hi": 45,
      "categories": {
        "reasoning": 65.7143,
        "math": 84.0,
        "coding": 86.0301,
        "instruction": 96.5,
        "knowledge": 96.6667,
        "long_context": 61.1111,
        "multilingual": 75.0,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8603,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.965,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3333,
          "delta_mean": 0.8,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3429,
          "delta_mean": 0.7143,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5417,
          "delta_mean": 1.2917,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.75,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 18.3536,
        "decode_tps_c4": 17.0396,
        "ttft_s": 0.7993,
        "prefill_tps_4k": 348.8265,
        "prefill_tps_12k": 346.3842,
        "load_seconds": 22.1702,
        "size_gb": 24.4226,
        "vram_gb": 24.4226,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "cogito-8b-v1-preview-llama",
      "display_name": "Cogito 8B V1 Preview Llama",
      "family": "llama",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Llama 3.1",
      "license": "llama3.2",
      "quantization": "Q4_K_M",
      "ollama_ref": "cogito:8b-v1-preview-llama-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 80,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 83.8743,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 7.1657,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.6617,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9802,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 2.8206,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 79.6,
      "ci_high": 87.77,
      "hard_score": 54.1229,
      "hard_n": 71,
      "ext_overall": 83.9566,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        117,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        126,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        146,
        148,
        149,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        168,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        200,
        202,
        215,
        216,
        221,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 78,
      "class_rank": 16,
      "rank_lo": 47,
      "rank_hi": 105,
      "class_rank_lo": 8,
      "class_rank_hi": 30,
      "categories": {
        "reasoning": 80.0,
        "math": 64.0,
        "coding": 56.0909,
        "instruction": 93.1111,
        "knowledge": 96.6667,
        "long_context": 83.3333,
        "multilingual": 81.6667,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5609,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5333,
          "delta_mean": 1.3333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9311,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6333,
          "delta_mean": 1.0667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6,
          "delta_mean": 0.5429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.1667,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.2234,
        "decode_tps_c4": 24.6189,
        "ttft_s": 0.3727,
        "prefill_tps_4k": 275.4079,
        "prefill_tps_12k": 226.1676,
        "load_seconds": 128.875,
        "size_gb": 22.351,
        "vram_gb": 22.351,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen3-4b-thinking-2507",
      "display_name": "Qwen3-4B-Thinking-2507",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 4.02,
      "arch": "dense",
      "hf_downloads": 558712,
      "hf_rank": 280,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:4b-thinking-2507-q4_K_M",
      "size_bytes": 2497280480,
      "model_version_id": 238,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 83.5194,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 5.7931,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8162,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8545,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 5.5283,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 80.2,
      "ci_high": 87.01,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 94.7394,
      "n_truncated": 45,
      "n_extended": 45,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        229,
        230,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 79,
      "class_rank": 17,
      "rank_lo": 47,
      "rank_hi": 111,
      "class_rank_lo": 7,
      "class_rank_hi": 31,
      "categories": {
        "reasoning": 74.2857,
        "math": 76.0,
        "coding": 26.6667,
        "instruction": 88.6111,
        "knowledge": 100.0,
        "long_context": 94.4444,
        "multilingual": 91.6667,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3333,
          "delta_mean": 0.3667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8861,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1667,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.9444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9143,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2917,
          "delta_mean": 1.4167,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 21.4766,
        "decode_tps_c4": 32.1159,
        "ttft_s": 0.2355,
        "prefill_tps_4k": 511.4064,
        "prefill_tps_12k": 352.3157,
        "load_seconds": 126.4797,
        "size_gb": 2.4973,
        "vram_gb": 2.4973,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "allenai-olmo-3-7b-instruct",
      "display_name": "Olmo-3-7B-Instruct",
      "family": "olmo",
      "org": "allenai",
      "params_b": 7.3,
      "arch": "dense",
      "hf_downloads": 476182,
      "hf_rank": 326,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "olmo-3:7b-instruct-q4_K_M",
      "size_bytes": 4471999712,
      "model_version_id": 1,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 82.8567,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 6.8176,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.8567,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8786,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 5.3654,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 8,
      "concurrency_text": "4 to 8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 78.8,
      "ci_high": 86.91,
      "hard_score": 43.3117,
      "hard_n": 38,
      "ext_overall": 83.936,
      "n_truncated": 11,
      "n_extended": 11,
      "n_gated": 1,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 80,
      "class_rank": 18,
      "rank_lo": 52,
      "rank_hi": 118,
      "class_rank_lo": 10,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 80.0,
        "math": 96.0,
        "coding": 57.7099,
        "instruction": 96.7222,
        "knowledge": 93.3333,
        "long_context": 44.4444,
        "multilingual": 86.6667,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5771,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0333,
          "delta_mean": 1.0667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9672,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.3667,
          "delta_mean": 0.7,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8,
          "delta_mean": 0.5429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7083,
          "delta_mean": 1.2083,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.2493,
        "decode_tps_c4": 22.758,
        "ttft_s": 0.6463,
        "prefill_tps_4k": 234.673,
        "prefill_tps_12k": 194.6145,
        "load_seconds": 141.3684,
        "size_gb": 14.3642,
        "vram_gb": 14.3642,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-3-4b-it",
      "display_name": "gemma-3-4b-it",
      "family": "gemma",
      "org": "google",
      "params_b": 4.3,
      "arch": "dense",
      "hf_downloads": 1237067,
      "hf_rank": 128,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3:4b-it-q4_K_M",
      "size_bytes": 3338792448,
      "model_version_id": 8,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 82.0911,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.6857,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3452,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0643,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 9.6925,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.59,
      "ci_high": 86.47,
      "hard_score": 48.2917,
      "hard_n": 38,
      "ext_overall": 83.1705,
      "n_truncated": 23,
      "n_extended": 19,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 81,
      "class_rank": 19,
      "rank_lo": 58,
      "rank_hi": 123,
      "class_rank_lo": 11,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 71.4286,
        "math": 84.0,
        "coding": 71.3917,
        "instruction": 85.8889,
        "knowledge": 80.0,
        "long_context": 61.1111,
        "multilingual": 100.0,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 75.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7139,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8589,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 1.3333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.84,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9429,
          "delta_mean": 0.5714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 16.9063,
        "decode_tps_c4": 38.6304,
        "ttft_s": 0.3071,
        "prefill_tps_4k": 492.6016,
        "prefill_tps_12k": 490.4903,
        "load_seconds": 27.1081,
        "size_gb": 4.4594,
        "vram_gb": 4.4594,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lgai-exaone-exaone-3-5-7-8b-instruct",
      "display_name": "EXAONE-3.5-7.8B-Instruct",
      "family": "exaone",
      "org": "LGAI-EXAONE",
      "params_b": 7.82,
      "arch": "dense",
      "hf_downloads": 300973,
      "hf_rank": 488,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "exaone-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "exaone3.5:7.8b-instruct-q4_K_M",
      "size_bytes": 4770650016,
      "model_version_id": 16,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 82.0459,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.7221,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.431,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0426,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 6.1264,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.56,
      "ci_high": 86.36,
      "hard_score": 43.0017,
      "hard_n": 38,
      "ext_overall": 82.0459,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 82,
      "class_rank": 20,
      "rank_lo": 59,
      "rank_hi": 122,
      "class_rank_lo": 11,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 80.0,
        "math": 72.0,
        "coding": 70.08,
        "instruction": 80.7778,
        "knowledge": 100.0,
        "long_context": 55.5556,
        "multilingual": 88.3333,
        "tool_use": 100.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7008,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6333,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8078,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8667,
          "delta_mean": 1.5333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5714,
          "delta_mean": 0.7714,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9583,
          "delta_mean": 1.2917,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.4804,
        "decode_tps_c4": 23.0027,
        "ttft_s": 0.3985,
        "prefill_tps_4k": 277.5624,
        "prefill_tps_12k": 228.5496,
        "load_seconds": 70.1358,
        "size_gb": 13.6406,
        "vram_gb": 13.6406,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-vl-7b-instruct",
      "display_name": "Qwen2.5-VL-7B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 8.29,
      "arch": "dense",
      "hf_downloads": 5670393,
      "hf_rank": 24,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5vl:7b-q4_K_M",
      "size_bytes": 5969233408,
      "model_version_id": 42,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 82.0256,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.4445,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9119,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0236,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 4.0605,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.75,
      "ci_high": 86.34,
      "hard_score": 32.124,
      "hard_n": 38,
      "ext_overall": 82.3431,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 83,
      "class_rank": 21,
      "rank_lo": 59,
      "rank_hi": 121,
      "class_rank_lo": 11,
      "class_rank_hi": 37,
      "categories": {
        "reasoning": 74.2857,
        "math": 68.0,
        "coding": 64.3895,
        "instruction": 78.7778,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 83.3333,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6439,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6,
          "delta_mean": 1.1333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7878,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 1.0,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9429,
          "delta_mean": 0.9429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 1.0417,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.125,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.2455,
        "decode_tps_c4": 28.35,
        "ttft_s": 0.3966,
        "prefill_tps_4k": 305.576,
        "prefill_tps_12k": 277.6815,
        "load_seconds": 63.9528,
        "size_gb": 13.2466,
        "vram_gb": 13.2466,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "command-r7b-arabic-7b-02-2025",
      "display_name": "Command R7B Arabic 7B 02 2025",
      "family": "cohere",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "cc-by-nc-4.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "command-r7b-arabic:7b-02-2025-q4_K_M",
      "size_bytes": 5100000000,
      "model_version_id": 81,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 81.9195,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.6798,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.1614,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0402,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 3.366,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.46,
      "ci_high": 86.55,
      "hard_score": 49.5261,
      "hard_n": 71,
      "ext_overall": 82.3639,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        43,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        94,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        166,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 84,
      "class_rank": 22,
      "rank_lo": 53,
      "rank_hi": 118,
      "class_rank_lo": 9,
      "class_rank_hi": 36,
      "categories": {
        "reasoning": 74.2857,
        "math": 68.0,
        "coding": 67.3784,
        "instruction": 84.8333,
        "knowledge": 93.3333,
        "long_context": 77.7778,
        "multilingual": 90.8333,
        "tool_use": 85.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6738,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4333,
          "delta_mean": 1.1,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8483,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4,
          "delta_mean": 1.4,
          "disputed": 9,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0571,
          "delta_mean": 0.7429,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.0833,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.9617,
        "decode_tps_c4": 16.1736,
        "ttft_s": 0.5112,
        "prefill_tps_4k": 248.2878,
        "prefill_tps_12k": 229.1643,
        "load_seconds": 92.39,
        "size_gb": 10.4904,
        "vram_gb": 10.4904,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-2-8b",
      "display_name": "Granite4.2 8B",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.2:8b-q4_K_M",
      "size_bytes": 5348000000,
      "model_version_id": 210,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 81.8653,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 7.0848,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 7.0178,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9975,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 3.7097,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.21,
      "ci_high": 86.29,
      "hard_score": 51.4437,
      "hard_n": 35,
      "ext_overall": 86.9076,
      "n_truncated": 60,
      "n_extended": 58,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        200,
        201,
        202,
        215,
        216,
        221,
        228,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 85,
      "class_rank": 23,
      "rank_lo": 59,
      "rank_hi": 128,
      "class_rank_lo": 10,
      "class_rank_hi": 42,
      "categories": {
        "reasoning": 65.7143,
        "math": 68.0,
        "coding": 85.1288,
        "instruction": 94.0556,
        "knowledge": 100.0,
        "long_context": 55.5556,
        "multilingual": 88.3333,
        "tool_use": 80.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8513,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9406,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8667,
          "delta_mean": 0.9333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7576,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2059,
          "delta_mean": 0.6875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 1.875,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.062,
        "decode_tps_c4": 24.8466,
        "ttft_s": 0.4154,
        "prefill_tps_4k": 243.6117,
        "prefill_tps_12k": 196.7583,
        "load_seconds": 169.1508,
        "size_gb": 27.1039,
        "vram_gb": 27.1039,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-r1-8b-llama-distill",
      "display_name": "Deepseek R1 8B Llama Distill",
      "family": "deepseek",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3.1",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:8b-llama-distill-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 89,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 81.7289,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 6.0312,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.7773,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9838,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 4.1807,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 76.92,
      "ci_high": 86.08,
      "hard_score": 57.0316,
      "hard_n": 71,
      "ext_overall": 84.4481,
      "n_truncated": 15,
      "n_extended": 15,
      "n_gated": 1,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        131,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        141,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 86,
      "class_rank": 24,
      "rank_lo": 56,
      "rank_hi": 122,
      "class_rank_lo": 11,
      "class_rank_hi": 37,
      "categories": {
        "reasoning": 91.4286,
        "math": 96.0,
        "coding": 52.798,
        "instruction": 82.2778,
        "knowledge": 90.0,
        "long_context": 72.2222,
        "multilingual": 71.6667,
        "tool_use": 100.0,
        "safety": 79.1667,
        "sycophancy": 100.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.528,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1,
          "delta_mean": 0.8333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8228,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5,
          "delta_mean": 1.0667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9714,
          "delta_mean": 0.8286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5652,
          "delta_mean": 1.5652,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.8841,
        "decode_tps_c4": 18.2193,
        "ttft_s": 0.3667,
        "prefill_tps_4k": 276.9593,
        "prefill_tps_12k": 230.6728,
        "load_seconds": 136.282,
        "size_gb": 22.351,
        "vram_gb": 22.351,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-4-mini-instruct",
      "display_name": "Phi-4-mini-instruct",
      "family": "phi",
      "org": "microsoft",
      "params_b": 3.84,
      "arch": "dense",
      "hf_downloads": 363242,
      "hf_rank": 423,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi4-mini:3.8b-q4_K_M",
      "size_bytes": 2491874624,
      "model_version_id": 24,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 81.5627,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.5548,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9938,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1214,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 16.6697,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 77.14,
      "ci_high": 85.95,
      "hard_score": 34.4797,
      "hard_n": 38,
      "ext_overall": 81.5627,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 1,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        34,
        38,
        41,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 87,
      "class_rank": 2,
      "rank_lo": 62,
      "rank_hi": 123,
      "class_rank_lo": 1,
      "class_rank_hi": 8,
      "categories": {
        "reasoning": 77.1429,
        "math": 76.0,
        "coding": 60.9773,
        "instruction": 84.1111,
        "knowledge": 96.6667,
        "long_context": 66.6667,
        "multilingual": 81.6667,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6098,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1667,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8411,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 1.4667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6857,
          "delta_mean": 1.0571,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2917,
          "delta_mean": 1.375,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.1375,
        "decode_tps_c4": 43.73,
        "ttft_s": 0.2254,
        "prefill_tps_4k": 514.2015,
        "prefill_tps_12k": 372.5218,
        "load_seconds": 73.2867,
        "size_gb": 11.3924,
        "vram_gb": 11.3924,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-10b-instruct",
      "display_name": "Falcon3 10B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 10.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3-tmplfix:10b-instruct-q4_K_M",
      "size_bytes": 6300000000,
      "model_version_id": 189,
      "run_id": "r20261028-bv1-7ce712",
      "battery_version": "v1",
      "tested_at": "2026-10-28T02:36:44.578Z",
      "overall": 81.2578,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 6.9505,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.4676,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1643,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 2.7614,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 76.55,
      "ci_high": 85.83,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 81.2578,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 88,
      "class_rank": 13,
      "rank_lo": 57,
      "rank_hi": 133,
      "class_rank_lo": 9,
      "class_rank_hi": 18,
      "categories": {
        "reasoning": 80.0,
        "math": 60.0,
        "coding": 64.3203,
        "instruction": 84.7778,
        "knowledge": 96.6667,
        "long_context": 72.2222,
        "multilingual": 83.3333,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 75.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6432,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3333,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8478,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4667,
          "delta_mean": 1.2667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3714,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9167,
          "delta_mean": 1.5,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.0354,
        "decode_tps_c4": 20.8677,
        "ttft_s": 0.5302,
        "prefill_tps_4k": 199.8057,
        "prefill_tps_12k": 173.6875,
        "load_seconds": 146.084,
        "size_gb": 6.2875,
        "vram_gb": 6.2875,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-llama-3-1-8b-instruct",
      "display_name": "Llama-3.1-8B-Instruct",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 8.03,
      "arch": "dense",
      "hf_downloads": 6104298,
      "hf_rank": 21,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3.1",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama3.1:8b-instruct-q4_K_M",
      "size_bytes": 4920738944,
      "model_version_id": 18,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 81.017,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.8374,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3348,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8488,
      "agreement": "high",
      "disputed": 4,
      "tokens_per_s": 5.7699,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 76.5,
      "ci_high": 85.17,
      "hard_score": 40.5675,
      "hard_n": 38,
      "ext_overall": 81.017,
      "n_truncated": 8,
      "n_extended": 8,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        19,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 89,
      "class_rank": 25,
      "rank_lo": 65,
      "rank_hi": 125,
      "class_rank_lo": 14,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 60.0,
        "math": 56.0,
        "coding": 65.5976,
        "instruction": 92.5556,
        "knowledge": 90.0,
        "long_context": 83.3333,
        "multilingual": 81.6667,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3333,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9256,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9,
          "delta_mean": 0.6667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8571,
          "delta_mean": 0.8857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 1.0417,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.551,
        "decode_tps_c4": 23.5598,
        "ttft_s": 0.406,
        "prefill_tps_4k": 278.2606,
        "prefill_tps_12k": 228.8605,
        "load_seconds": 70.082,
        "size_gb": 13.7275,
        "vram_gb": 13.7275,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "ministral-3-3b-instruct-2512",
      "display_name": "Ministral 3 3B Instruct 2512",
      "family": "mistral",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "ministral-3:3b-instruct-2512-q4_K_M",
      "size_bytes": 3000000000,
      "model_version_id": 138,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 80.9089,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 6.2655,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.6907,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2114,
      "agreement": "high",
      "disputed": 21,
      "tokens_per_s": 23.2956,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 76.49,
      "ci_high": 85.18,
      "hard_score": 50.3443,
      "hard_n": 71,
      "ext_overall": 81.2263,
      "n_truncated": 12,
      "n_extended": 12,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        132,
        133,
        134,
        135,
        136,
        137,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        167,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 90,
      "class_rank": 3,
      "rank_lo": 61,
      "rank_hi": 124,
      "class_rank_lo": 1,
      "class_rank_hi": 11,
      "categories": {
        "reasoning": 74.2857,
        "math": 88.0,
        "coding": 60.6165,
        "instruction": 67.5,
        "knowledge": 93.3333,
        "long_context": 61.1111,
        "multilingual": 92.5,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6062,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9333,
          "delta_mean": 0.8667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.675,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2667,
          "delta_mean": 1.8333,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6286,
          "delta_mean": 0.8571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.375,
          "delta_mean": 1.375,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 70.8715,
        "decode_tps_c4": 156.7907,
        "ttft_s": 0.0608,
        "prefill_tps_4k": 2515.9526,
        "prefill_tps_12k": 1632.0885,
        "load_seconds": 75.094,
        "size_gb": 16.5584,
        "vram_gb": 16.5584,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "ibm-granite-granite-4-1-3b",
      "display_name": "granite-4.1-3b",
      "family": "granite",
      "org": "ibm-granite",
      "params_b": 3.4,
      "arch": "dense",
      "hf_downloads": 540575,
      "hf_rank": 292,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.1:3b-q4_K_M",
      "size_bytes": 2099501664,
      "model_version_id": 15,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 80.7146,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.5995,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3569,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0274,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 10.1824,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 75.88,
      "ci_high": 85.26,
      "hard_score": 42.0138,
      "hard_n": 38,
      "ext_overall": 80.7146,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        2,
        4,
        13,
        17,
        20,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 91,
      "class_rank": 4,
      "rank_lo": 66,
      "rank_hi": 129,
      "class_rank_lo": 1,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 65.7143,
        "math": 80.0,
        "coding": 62.6063,
        "instruction": 93.9444,
        "knowledge": 90.0,
        "long_context": 66.6667,
        "multilingual": 86.6667,
        "tool_use": 85.0,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6261,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2667,
          "delta_mean": 1.1667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9394,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 0.9667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1429,
          "delta_mean": 0.6286,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7083,
          "delta_mean": 1.375,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.1362,
        "decode_tps_c4": 49.0449,
        "ttft_s": 0.246,
        "prefill_tps_4k": 472.0635,
        "prefill_tps_12k": 345.8545,
        "load_seconds": 48.167,
        "size_gb": 7.796,
        "vram_gb": 7.796,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-gemma-2-9b-it",
      "display_name": "gemma-2-9b-it",
      "family": "gemma",
      "org": "google",
      "params_b": 9.24,
      "arch": "dense",
      "hf_downloads": 1026916,
      "hf_rank": 167,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma2:9b-instruct-q4_K_M",
      "size_bytes": 5761058432,
      "model_version_id": 5,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 79.8919,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.9117,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3893,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0119,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 2.2521,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 75.5,
      "ci_high": 84.19,
      "hard_score": 36.8431,
      "hard_n": 38,
      "ext_overall": 79.8919,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        4,
        13,
        17,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 92,
      "class_rank": 14,
      "rank_lo": 69,
      "rank_hi": 131,
      "class_rank_lo": 12,
      "class_rank_hi": 18,
      "categories": {
        "reasoning": 68.5714,
        "math": 60.0,
        "coding": 51.7892,
        "instruction": 92.5556,
        "knowledge": 90.0,
        "long_context": 61.1111,
        "multilingual": 100.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5179,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9256,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6,
          "delta_mean": 0.8667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3714,
          "delta_mean": 1.1429,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.375,
          "delta_mean": 0.9583,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 7.5773,
        "decode_tps_c4": 11.7128,
        "ttft_s": 0.6862,
        "prefill_tps_4k": 171.6401,
        "prefill_tps_12k": 171.3629,
        "load_seconds": 154.3646,
        "size_gb": 14.7141,
        "vram_gb": 14.7141,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma3n-e2b-it",
      "display_name": "Gemma3n E2B IT",
      "family": "gemma",
      "org": null,
      "params_b": 5.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 2.0,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3n:e2b-it-q4_K_M",
      "size_bytes": 5600000000,
      "model_version_id": 104,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 79.8536,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 6.5693,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.1374,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2405,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 6.8631,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 75.22,
      "ci_high": 84.66,
      "hard_score": 55.7108,
      "hard_n": 71,
      "ext_overall": 81.0198,
      "n_truncated": 29,
      "n_extended": 29,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        83,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        132,
        133,
        134,
        135,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 93,
      "class_rank": 26,
      "rank_lo": 64,
      "rank_hi": 127,
      "class_rank_lo": 13,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 68.5714,
        "math": 76.0,
        "coding": 61.3886,
        "instruction": 89.6667,
        "knowledge": 83.3333,
        "long_context": 72.2222,
        "multilingual": 80.0,
        "tool_use": 87.5,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6139,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7,
          "delta_mean": 1.0333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8967,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7667,
          "delta_mean": 1.5667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4286,
          "delta_mean": 0.6857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.625,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.816,
        "decode_tps_c4": 26.0991,
        "ttft_s": 0.3533,
        "prefill_tps_4k": 483.8795,
        "prefill_tps_12k": 486.1134,
        "load_seconds": 19.593,
        "size_gb": 3.3972,
        "vram_gb": 3.3972,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-ai-deepseek-r1-0528-qwen3-8b",
      "display_name": "DeepSeek-R1-0528-Qwen3-8B",
      "family": "deepseek",
      "org": "deepseek-ai",
      "params_b": 8.19,
      "arch": "dense",
      "hf_downloads": 682235,
      "hf_rank": 229,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:8b-0528-qwen3-q4_K_M",
      "size_bytes": 5225373760,
      "model_version_id": 230,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 79.554,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 6.4112,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9088,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8971,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 3.6596,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 75.48,
      "ci_high": 83.86,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 86.5543,
      "n_truncated": 40,
      "n_extended": 40,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        221,
        233,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 94,
      "class_rank": 27,
      "rank_lo": 67,
      "rank_hi": 135,
      "class_rank_lo": 14,
      "class_rank_hi": 43,
      "categories": {
        "reasoning": 77.1429,
        "math": 48.0,
        "coding": 42.1212,
        "instruction": 84.2778,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 95.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4212,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2667,
          "delta_mean": 0.2667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8428,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 1.5333,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8857,
          "delta_mean": 0.6857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7083,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.1974,
        "decode_tps_c4": 18.3877,
        "ttft_s": 0.3887,
        "prefill_tps_4k": 289.4775,
        "prefill_tps_12k": 229.5355,
        "load_seconds": 135.4361,
        "size_gb": 5.2254,
        "vram_gb": 5.2254,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-3-8b",
      "display_name": "Granite3.3 8B",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.3:8b",
      "size_bytes": 4900000000,
      "model_version_id": 117,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 79.395,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.5545,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.0483,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0133,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 8.5991,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 74.74,
      "ci_high": 83.9,
      "hard_score": 40.2009,
      "hard_n": 71,
      "ext_overall": 79.395,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        115,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        125,
        128,
        130,
        133,
        134,
        135,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 95,
      "class_rank": 28,
      "rank_lo": 66,
      "rank_hi": 129,
      "class_rank_lo": 14,
      "class_rank_hi": 38,
      "categories": {
        "reasoning": 62.8571,
        "math": 76.0,
        "coding": 56.642,
        "instruction": 85.7222,
        "knowledge": 96.6667,
        "long_context": 66.6667,
        "multilingual": 72.5,
        "tool_use": 97.5,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5664,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8572,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8333,
          "delta_mean": 1.1333,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.725,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5833,
          "delta_mean": 1.3333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.975,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.3445,
        "decode_tps_c4": 25.8932,
        "ttft_s": 0.4756,
        "prefill_tps_4k": 254.4203,
        "prefill_tps_12k": 203.4872,
        "load_seconds": 154.485,
        "size_gb": 26.9319,
        "vram_gb": 26.9319,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-tiny-h",
      "display_name": "Granite4 Tiny H",
      "family": "granite",
      "org": null,
      "params_b": 7.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 1.0,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4:tiny-h",
      "size_bytes": 4200000000,
      "model_version_id": 127,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 79.1049,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.9543,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.3171,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1719,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 4.494,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 74.13,
      "ci_high": 84.18,
      "hard_score": 54.3157,
      "hard_n": 71,
      "ext_overall": 79.1049,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        71,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        107,
        108,
        109,
        110,
        111,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        183,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 96,
      "class_rank": 29,
      "rank_lo": 64,
      "rank_hi": 132,
      "class_rank_lo": 13,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 74.2857,
        "math": 68.0,
        "coding": 80.7694,
        "instruction": 79.7222,
        "knowledge": 86.6667,
        "long_context": 66.6667,
        "multilingual": 80.0,
        "tool_use": 80.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8077,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 1.0333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7972,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.6667,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8857,
          "delta_mean": 0.7429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5417,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 24.2409,
        "decode_tps_c4": 40.9338,
        "ttft_s": 0.4229,
        "prefill_tps_4k": 442.007,
        "prefill_tps_12k": 476.0533,
        "load_seconds": 19.359,
        "size_gb": 6.0593,
        "vram_gb": 6.0593,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-ai-deepseek-coder-v2-lite-instruct",
      "display_name": "DeepSeek-Coder-V2-Lite-Instruct",
      "family": "deepseek",
      "org": "deepseek-ai",
      "params_b": 15.71,
      "arch": "moe",
      "hf_downloads": 895419,
      "hf_rank": 188,
      "active_params_b": 2.4,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-coder-v2:16b-lite-instruct-q4_K_M",
      "size_bytes": 10364416480,
      "model_version_id": 3,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 78.9798,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 6.2429,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.78,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9629,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 0.4924,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 74.33,
      "ci_high": 83.48,
      "hard_score": 38.9858,
      "hard_n": 38,
      "ext_overall": 78.9798,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        4,
        17,
        21,
        22,
        27,
        29,
        38,
        41,
        43,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 97,
      "class_rank": 15,
      "rank_lo": 69,
      "rank_hi": 136,
      "class_rank_lo": 12,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 65.7143,
        "math": 56.0,
        "coding": 72.5487,
        "instruction": 77.9444,
        "knowledge": 83.3333,
        "long_context": 77.7778,
        "multilingual": 86.6667,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7255,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7667,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7794,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.4667,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8333,
          "delta_mean": 1.2083,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.0589,
        "decode_tps_c4": 4.5617,
        "ttft_s": 0.5033,
        "prefill_tps_4k": 159.8893,
        "prefill_tps_12k": 110.0963,
        "load_seconds": 283.9216,
        "size_gb": 46.8006,
        "vram_gb": 46.8006,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "aya-expanse-32b",
      "display_name": "Aya Expanse 32B",
      "family": "cohere",
      "org": null,
      "params_b": 32.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "aya-expanse:32b-q4_K_M",
      "size_bytes": 19801000000,
      "model_version_id": 196,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 78.7594,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.5319,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9983,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0031,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 1.1714,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 74.62,
      "ci_high": 82.87,
      "hard_score": 37.6623,
      "hard_n": 35,
      "ext_overall": 78.7594,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        143,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 98,
      "class_rank": 42,
      "rank_lo": 72,
      "rank_hi": 137,
      "class_rank_lo": 41,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 54.2857,
        "math": 48.0,
        "coding": 64.7155,
        "instruction": 81.2778,
        "knowledge": 100.0,
        "long_context": 72.2222,
        "multilingual": 93.3333,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6472,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 1.0667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8128,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7667,
          "delta_mean": 1.7,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.4167,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.9938,
        "decode_tps_c4": 13.5274,
        "ttft_s": 0.4201,
        "prefill_tps_4k": 281.2772,
        "prefill_tps_12k": 88.1637,
        "load_seconds": 89.1578,
        "size_gb": 31.4334,
        "vram_gb": 31.4334,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-2-3b",
      "display_name": "Granite4.2 3B",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4.2:3b-q4_K_M",
      "size_bytes": 2244000000,
      "model_version_id": 209,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 78.2385,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 6.5334,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.5171,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9646,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 7.2141,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 73.69,
      "ci_high": 83.08,
      "hard_score": 36.2747,
      "hard_n": 35,
      "ext_overall": 83.8893,
      "n_truncated": 46,
      "n_extended": 45,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        215,
        216,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 99,
      "class_rank": 5,
      "rank_lo": 72,
      "rank_hi": 140,
      "class_rank_lo": 2,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 74.2857,
        "math": 56.0,
        "coding": 61.5831,
        "instruction": 91.1667,
        "knowledge": 90.0,
        "long_context": 61.1111,
        "multilingual": 75.0,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6158,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3667,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9117,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5333,
          "delta_mean": 0.9333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0588,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6857,
          "delta_mean": 0.7647,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.7917,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 24.2936,
        "decode_tps_c4": 50.144,
        "ttft_s": 0.1966,
        "prefill_tps_4k": 472.0694,
        "prefill_tps_12k": 346.6334,
        "load_seconds": 84.5551,
        "size_gb": 13.1857,
        "vram_gb": 13.1857,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "huggingfacetb-smollm3-3b",
      "display_name": "SmolLM3-3B",
      "family": "smollm",
      "org": "HuggingFaceTB",
      "params_b": 3.08,
      "arch": "dense",
      "hf_downloads": 628733,
      "hf_rank": 247,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/unsloth/SmolLM3-3B-GGUF:Q4_K_M",
      "size_bytes": null,
      "model_version_id": 14,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 78.0448,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.2919,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.854,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0555,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 8.0249,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 72.95,
      "ci_high": 83.08,
      "hard_score": 32.6834,
      "hard_n": 38,
      "ext_overall": 78.0448,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 1,
      "beats": [
        4,
        17,
        21,
        22,
        27,
        29,
        41,
        43,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 100,
      "class_rank": 6,
      "rank_lo": 71,
      "rank_hi": 142,
      "class_rank_lo": 1,
      "class_rank_hi": 14,
      "categories": {
        "reasoning": 71.4286,
        "math": 72.0,
        "coding": 57.697,
        "instruction": 86.0,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 75.0,
        "tool_use": 82.5,
        "safety": 83.3333,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.577,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8667,
          "delta_mean": 0.9,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.86,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7333,
          "delta_mean": 1.2,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6286,
          "delta_mean": 0.8857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0417,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 22.2367,
        "decode_tps_c4": 50.7828,
        "ttft_s": 0.1972,
        "prefill_tps_4k": 684.0474,
        "prefill_tps_12k": 594.8398,
        "load_seconds": 75.7615,
        "size_gb": 11.9592,
        "vram_gb": 11.9592,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "phi3-14b-medium-128k-instruct",
      "display_name": "Phi3 14B Medium 128K Instruct",
      "family": "phi",
      "org": null,
      "params_b": 14.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi3:14b-medium-128k-instruct-q4_K_M",
      "size_bytes": 8600000000,
      "model_version_id": 149,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 77.9607,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.4114,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8031,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.4155,
      "agreement": "high",
      "disputed": 22,
      "tokens_per_s": 2.1102,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 73.23,
      "ci_high": 82.55,
      "hard_score": 49.6024,
      "hard_n": 71,
      "ext_overall": 78.4237,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 101,
      "class_rank": 16,
      "rank_lo": 67,
      "rank_hi": 136,
      "class_rank_lo": 11,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 77.1429,
        "math": 56.0,
        "coding": 61.3923,
        "instruction": 71.5556,
        "knowledge": 90.0,
        "long_context": 88.8889,
        "multilingual": 69.1667,
        "tool_use": 87.5,
        "safety": 100.0,
        "sycophancy": 66.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6139,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7156,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7333,
          "delta_mean": 1.7667,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8571,
          "delta_mean": 1.0857,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.5,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.625,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": null,
        "decode_tps_c4": 17.0354,
        "ttft_s": null,
        "prefill_tps_4k": null,
        "prefill_tps_12k": 126.4916,
        "load_seconds": null,
        "size_gb": null,
        "vram_gb": null,
        "gpu_fraction": null,
        "context_length": null
      },
      "perf_units": {
        "decode_tps_c4": "tok/s",
        "prefill_tps_12k": "tok/s"
      }
    },
    {
      "slug": "allenai-olmo-3-7b-think",
      "display_name": "Olmo-3-7B-Think",
      "family": "olmo",
      "org": "allenai",
      "params_b": 7.3,
      "arch": "dense",
      "hf_downloads": 452109,
      "hf_rank": 339,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "olmo-3:7b-think-q4_K_M",
      "size_bytes": 4471999712,
      "model_version_id": 229,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 77.7496,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.9164,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.0243,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8488,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 2.6085,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 73.06,
      "ci_high": 82.46,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 88.3535,
      "n_truncated": 56,
      "n_extended": 56,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 102,
      "class_rank": 30,
      "rank_lo": 68,
      "rank_hi": 146,
      "class_rank_lo": 14,
      "class_rank_hi": 48,
      "categories": {
        "reasoning": 62.8571,
        "math": 48.0,
        "coding": 40.0,
        "instruction": 98.3333,
        "knowledge": 100.0,
        "long_context": 88.8889,
        "multilingual": 90.0,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 66.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6333,
          "delta_mean": 0.3333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 9.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.7,
          "delta_mean": 0.4333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3714,
          "delta_mean": 0.6857,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.7917,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.4627,
        "decode_tps_c4": 23.9585,
        "ttft_s": 0.6121,
        "prefill_tps_4k": 256.7173,
        "prefill_tps_12k": 229.1374,
        "load_seconds": 209.8619,
        "size_gb": 4.472,
        "vram_gb": 4.472,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen3-vl-2b-instruct",
      "display_name": "Qwen3 VL 2B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:2b-instruct-q4_K_M",
      "size_bytes": 1900000000,
      "model_version_id": 168,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 77.364,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.2388,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.54,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1431,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 26.6407,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 72.49,
      "ci_high": 82.07,
      "hard_score": 43.9696,
      "hard_n": 71,
      "ext_overall": 80.4733,
      "n_truncated": 34,
      "n_extended": 33,
      "n_gated": 1,
      "beats": [
        17,
        21,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 103,
      "class_rank": 1,
      "rank_lo": 68,
      "rank_hi": 137,
      "class_rank_lo": 1,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 65.7143,
        "math": 64.0,
        "coding": 49.2281,
        "instruction": 89.8333,
        "knowledge": 93.3333,
        "long_context": 66.6667,
        "multilingual": 75.8333,
        "tool_use": 100.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4923,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2667,
          "delta_mean": 1.0667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8983,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4,
          "delta_mean": 0.8333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2,
          "delta_mean": 0.8571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2083,
          "delta_mean": 1.3333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 26.4991,
        "decode_tps_c4": 26.2147,
        "ttft_s": 0.145,
        "prefill_tps_4k": 933.3264,
        "prefill_tps_12k": 787.9321,
        "load_seconds": 16.532,
        "size_gb": 3.5949,
        "vram_gb": 3.5949,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-3b-instruct",
      "display_name": "Qwen2.5-3B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 3.09,
      "arch": "dense",
      "hf_downloads": 4213957,
      "hf_rank": 34,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "qwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:3b-instruct-q4_K_M",
      "size_bytes": 1929903008,
      "model_version_id": 34,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 76.7703,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 6.0819,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.6483,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9383,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 12.124,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 8,
      "concurrency_text": "4 to 8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 72.04,
      "ci_high": 81.57,
      "hard_score": 25.352,
      "hard_n": 38,
      "ext_overall": 77.0018,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 104,
      "class_rank": 7,
      "rank_lo": 76,
      "rank_hi": 148,
      "class_rank_lo": 2,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 71.4286,
        "math": 72.0,
        "coding": 48.7262,
        "instruction": 82.6667,
        "knowledge": 90.0,
        "long_context": 61.1111,
        "multilingual": 75.0,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4873,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8333,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8267,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 1.3,
          "disputed": 10,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.2917,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 21.8785,
        "decode_tps_c4": 44.9175,
        "ttft_s": 0.2039,
        "prefill_tps_4k": 680.6535,
        "prefill_tps_12k": 591.14,
        "load_seconds": 23.3712,
        "size_gb": 4.7279,
        "vram_gb": 4.7279,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "openthinker-7b-v2",
      "display_name": "Openthinker 7B V2",
      "family": "other",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2.5",
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "openthinker:7b-v2-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 146,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 76.7682,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 5.2676,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.5279,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1925,
      "agreement": "high",
      "disputed": 21,
      "tokens_per_s": 4.4956,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 72.09,
      "ci_high": 81.6,
      "hard_score": 54.1236,
      "hard_n": 71,
      "ext_overall": 82.9615,
      "n_truncated": 36,
      "n_extended": 36,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        151,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 105,
      "class_rank": 31,
      "rank_lo": 70,
      "rank_hi": 140,
      "class_rank_lo": 15,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 88.5714,
        "math": 88.0,
        "coding": 52.5926,
        "instruction": 62.4444,
        "knowledge": 100.0,
        "long_context": 55.5556,
        "multilingual": 87.5,
        "tool_use": 75.0,
        "safety": 81.25,
        "sycophancy": 83.3333,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5259,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5,
          "delta_mean": 1.6333,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6244,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 1.2333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5882,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5143,
          "delta_mean": 0.4706,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8125,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.375,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.5576,
        "decode_tps_c4": 30.0761,
        "ttft_s": 0.3586,
        "prefill_tps_4k": 307.5555,
        "prefill_tps_12k": 280.2105,
        "load_seconds": 59.593,
        "size_gb": 12.5369,
        "vram_gb": 12.5369,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-micro-h",
      "display_name": "Granite4 Micro H",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4:micro-h",
      "size_bytes": 1900000000,
      "model_version_id": 126,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 76.6382,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.6895,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.2393,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9664,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 3.7941,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 72.03,
      "ci_high": 81.36,
      "hard_score": 46.9239,
      "hard_n": 71,
      "ext_overall": 76.6382,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        135,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 106,
      "class_rank": 8,
      "rank_lo": 72,
      "rank_hi": 139,
      "class_rank_lo": 1,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 62.8571,
        "math": 60.0,
        "coding": 68.8309,
        "instruction": 90.0,
        "knowledge": 90.0,
        "long_context": 38.8889,
        "multilingual": 83.3333,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6883,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3333,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4,
          "delta_mean": 0.8,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1714,
          "delta_mean": 0.8571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 1.3333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.3735,
        "decode_tps_c4": 30.9272,
        "ttft_s": 0.6006,
        "prefill_tps_4k": 364.4477,
        "prefill_tps_12k": 369.9609,
        "load_seconds": 23.938,
        "size_gb": 3.9841,
        "vram_gb": 3.9841,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "exaone3-5-2-4b-instruct",
      "display_name": "Exaone3.5 2.4B Instruct",
      "family": "exaone",
      "org": null,
      "params_b": 2.4,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "exaone-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "exaone3.5:2.4b-instruct-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 94,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 76.256,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.4048,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9843,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2429,
      "agreement": "high",
      "disputed": 25,
      "tokens_per_s": 10.6826,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.94,
      "ci_high": 81.28,
      "hard_score": 47.2779,
      "hard_n": 71,
      "ext_overall": 76.256,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 107,
      "class_rank": 2,
      "rank_lo": 73,
      "rank_hi": 141,
      "class_rank_lo": 1,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 82.8571,
        "math": 76.0,
        "coding": 64.7244,
        "instruction": 73.5556,
        "knowledge": 90.0,
        "long_context": 50.0,
        "multilingual": 62.5,
        "tool_use": 95.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6472,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4333,
          "delta_mean": 0.9333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7356,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7333,
          "delta_mean": 1.8667,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1714,
          "delta_mean": 0.9143,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 1.625,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 100.8843,
        "decode_tps_c4": 256.0,
        "ttft_s": 0.0402,
        "prefill_tps_4k": 2889.9137,
        "prefill_tps_12k": 2324.5215,
        "load_seconds": 74.297,
        "size_gb": 11.8868,
        "vram_gb": 11.8868,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-7b-instruct",
      "display_name": "Falcon3 7B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3-tmplfix:7b-instruct-q4_K_M",
      "size_bytes": 4600000000,
      "model_version_id": 192,
      "run_id": "r20261028-bv1-7ce712",
      "battery_version": "v1",
      "tested_at": "2026-10-28T02:36:44.578Z",
      "overall": 76.2343,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.6186,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.064,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9817,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.9732,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 71.64,
      "ci_high": 81.28,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 76.2343,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        194,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 108,
      "class_rank": 32,
      "rank_lo": 71,
      "rank_hi": 150,
      "class_rank_lo": 15,
      "class_rank_hi": 50,
      "categories": {
        "reasoning": 71.4286,
        "math": 76.0,
        "coding": 58.7354,
        "instruction": 80.2222,
        "knowledge": 93.3333,
        "long_context": 55.5556,
        "multilingual": 67.5,
        "tool_use": 87.5,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5874,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.1333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8022,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9,
          "delta_mean": 1.3,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.675,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0286,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.875,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.507,
        "decode_tps_c4": 28.4187,
        "ttft_s": 0.3699,
        "prefill_tps_4k": 290.0792,
        "prefill_tps_12k": 249.6633,
        "load_seconds": 103.4836,
        "size_gb": 4.5697,
        "vram_gb": 4.5697,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen3-1-7b",
      "display_name": "Qwen3 1.7B",
      "family": "qwen",
      "org": null,
      "params_b": 1.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:1.7b-q4_K_M",
      "size_bytes": 1400000000,
      "model_version_id": 166,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 76.0225,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 5.91,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.45,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0257,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 38.3251,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 71.26,
      "ci_high": 80.87,
      "hard_score": 42.7946,
      "hard_n": 71,
      "ext_overall": 75.7976,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 1,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        191,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 109,
      "class_rank": 3,
      "rank_lo": 77,
      "rank_hi": 141,
      "class_rank_lo": 1,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 68.5714,
        "math": 72.0,
        "coding": 45.4642,
        "instruction": 86.7778,
        "knowledge": 76.6667,
        "long_context": 55.5556,
        "multilingual": 87.5,
        "tool_use": 100.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4546,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7667,
          "delta_mean": 0.9333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8678,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6667,
          "delta_mean": 1.3667,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4,
          "delta_mean": 0.8286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9167,
          "delta_mean": 1.375,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 38.5041,
        "decode_tps_c4": 43.4559,
        "ttft_s": 0.1264,
        "prefill_tps_4k": 1056.7413,
        "prefill_tps_12k": 779.2915,
        "load_seconds": 95.4523,
        "size_gb": 16.4959,
        "vram_gb": 16.4959,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistral-nemo-12b-instruct-2407",
      "display_name": "Mistral Nemo 12B Instruct 2407",
      "family": "mistral",
      "org": null,
      "params_b": 12.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral-nemo:12b-instruct-2407-q4_K_M",
      "size_bytes": 7500000000,
      "model_version_id": 141,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 75.9002,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 2,
      "judge_a_mean": 5.9388,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.6039,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0912,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 2.4837,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.55,
      "ci_high": 80.77,
      "hard_score": 37.6585,
      "hard_n": 71,
      "ext_overall": 75.9002,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        79,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        185,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 110,
      "class_rank": 17,
      "rank_lo": 78,
      "rank_hi": 142,
      "class_rank_lo": 13,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 57.1429,
        "math": 56.0,
        "coding": 57.2918,
        "instruction": 80.7222,
        "knowledge": 96.6667,
        "long_context": 77.7778,
        "multilingual": 85.0,
        "tool_use": 85.0,
        "safety": 87.5,
        "sycophancy": 75.0,
        "gotchas": 86.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5729,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.2667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8072,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7,
          "delta_mean": 1.4667,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7429,
          "delta_mean": 0.8857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4348,
          "delta_mean": 1.087,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.375,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.1804,
        "decode_tps_c4": 12.5021,
        "ttft_s": 0.5436,
        "prefill_tps_4k": 176.6196,
        "prefill_tps_12k": 152.6864,
        "load_seconds": 143.516,
        "size_gb": 29.1539,
        "vram_gb": 29.1539,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-llama-3-2-3b-instruct",
      "display_name": "Llama-3.2-3B-Instruct",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 3.21,
      "arch": "dense",
      "hf_downloads": 1628863,
      "hf_rank": 90,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3.2",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama3.2:3b-instruct-q4_K_M",
      "size_bytes": 2019377376,
      "model_version_id": 19,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 75.7368,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.5045,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.0957,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0274,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 10.744,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.98,
      "ci_high": 80.43,
      "hard_score": 33.5237,
      "hard_n": 38,
      "ext_overall": 75.7368,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 111,
      "class_rank": 9,
      "rank_lo": 84,
      "rank_hi": 150,
      "class_rank_lo": 3,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 68.5714,
        "math": 52.0,
        "coding": 52.9489,
        "instruction": 91.7222,
        "knowledge": 86.6667,
        "long_context": 55.5556,
        "multilingual": 81.6667,
        "tool_use": 92.5,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5295,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4667,
          "delta_mean": 1.1667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9172,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 1.1,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4286,
          "delta_mean": 0.8286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0,
          "delta_mean": 1.2917,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.6795,
        "decode_tps_c4": 49.7549,
        "ttft_s": 0.2033,
        "prefill_tps_4k": 497.209,
        "prefill_tps_12k": 379.5428,
        "load_seconds": 67.8224,
        "size_gb": 9.8717,
        "vram_gb": 9.8717,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lfm2-5-8b-a1b",
      "display_name": "Lfm2.5 8B A1B",
      "family": "lfm",
      "org": null,
      "params_b": 8.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 1.0,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "lfm2.5:8b-a1b-q4_K_M",
      "size_bytes": 5200000000,
      "model_version_id": 131,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 75.7244,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.8343,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4138,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0948,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 43.8586,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.75,
      "ci_high": 80.47,
      "hard_score": 54.2371,
      "hard_n": 71,
      "ext_overall": 78.8302,
      "n_truncated": 52,
      "n_extended": 52,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 112,
      "class_rank": 33,
      "rank_lo": 80,
      "rank_hi": 143,
      "class_rank_lo": 19,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 85.7143,
        "math": 88.0,
        "coding": 47.25,
        "instruction": 91.1111,
        "knowledge": 96.6667,
        "long_context": 61.1111,
        "multilingual": 81.6667,
        "tool_use": 30.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4725,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3333,
          "delta_mean": 1.2333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8857,
          "delta_mean": 1.0571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.375,
          "delta_mean": 1.3333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 43.4376,
        "decode_tps_c4": 41.6108,
        "ttft_s": 0.2304,
        "prefill_tps_4k": 941.7579,
        "prefill_tps_12k": 910.1233,
        "load_seconds": 9.5,
        "size_gb": 5.5042,
        "vram_gb": 5.5042,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "yi-34b-chat-v1-5",
      "display_name": "YI 34B Chat V1.5",
      "family": "yi",
      "org": null,
      "params_b": 34.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "yi:34b-chat-v1.5-q4_K_M",
      "size_bytes": 20659000000,
      "model_version_id": 228,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 75.2062,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.2226,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8807,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0881,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 1.0508,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.66,
      "ci_high": 79.88,
      "hard_score": 35.3664,
      "hard_n": 35,
      "ext_overall": 75.2062,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        191,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 113,
      "class_rank": 43,
      "rank_lo": 79,
      "rank_hi": 152,
      "class_rank_lo": 41,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 71.4286,
        "math": 60.0,
        "coding": 50.8164,
        "instruction": 76.5556,
        "knowledge": 100.0,
        "long_context": 38.8889,
        "multilingual": 83.3333,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5082,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2,
          "delta_mean": 0.9333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 1.6,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0286,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 1.25,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.004,
        "decode_tps_c4": 7.1484,
        "ttft_s": 1.6171,
        "prefill_tps_4k": 56.8029,
        "prefill_tps_12k": 59.1703,
        "load_seconds": 71.6967,
        "size_gb": 28.8297,
        "vram_gb": 28.8297,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "command-r-35b-08-2024",
      "display_name": "Command R 35B 08 2024",
      "family": "cohere",
      "org": null,
      "params_b": 35.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "command-r:35b-08-2024-q4_K_M",
      "size_bytes": 19801000000,
      "model_version_id": 201,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 75.1743,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 6.4457,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.876,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1188,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 0.9184,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.61,
      "ci_high": 79.75,
      "hard_score": 40.7179,
      "hard_n": 35,
      "ext_overall": 75.1743,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 114,
      "class_rank": 44,
      "rank_lo": 79,
      "rank_hi": 153,
      "class_rank_lo": 41,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 48.5714,
        "math": 32.0,
        "coding": 63.8864,
        "instruction": 84.0556,
        "knowledge": 93.3333,
        "long_context": 88.8889,
        "multilingual": 84.1667,
        "tool_use": 90.0,
        "safety": 91.6667,
        "sycophancy": 83.3333,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6389,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9667,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8406,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 1.3,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.32,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7714,
          "delta_mean": 0.8857,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9583,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.2315,
        "decode_tps_c4": 7.6874,
        "ttft_s": 1.3449,
        "prefill_tps_4k": 60.9149,
        "prefill_tps_12k": 55.0572,
        "load_seconds": 158.9786,
        "size_gb": 42.2044,
        "vram_gb": 42.2044,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen3-5-2b",
      "display_name": "Qwen3.5 2B",
      "family": "qwen",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3.5:2b-q4_K_M",
      "size_bytes": 1900000000,
      "model_version_id": 167,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 75.0878,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.0995,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.7133,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1086,
      "agreement": "high",
      "disputed": 18,
      "tokens_per_s": 33.4389,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 69.67,
      "ci_high": 80.26,
      "hard_score": 44.1214,
      "hard_n": 71,
      "ext_overall": 76.22,
      "n_truncated": 19,
      "n_extended": 19,
      "n_gated": 1,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 115,
      "class_rank": 4,
      "rank_lo": 81,
      "rank_hi": 144,
      "class_rank_lo": 1,
      "class_rank_hi": 7,
      "categories": {
        "reasoning": 74.2857,
        "math": 80.0,
        "coding": 52.5041,
        "instruction": 82.0556,
        "knowledge": 80.0,
        "long_context": 77.7778,
        "multilingual": 63.3333,
        "tool_use": 70.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.525,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4,
          "delta_mean": 0.8667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8206,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0333,
          "delta_mean": 1.4333,
          "disputed": 10,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8,
          "delta_mean": 0.7429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8333,
          "delta_mean": 1.75,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 33.3131,
        "decode_tps_c4": 32.1578,
        "ttft_s": 0.2042,
        "prefill_tps_4k": 961.3919,
        "prefill_tps_12k": 939.157,
        "load_seconds": 8.297,
        "size_gb": 1.8618,
        "vram_gb": 1.8618,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "coherelabs-c4ai-command-r7b-12-2024",
      "display_name": "Command R7B (12-2024)",
      "family": "cohere",
      "org": "CohereLabs",
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "cc-by-nc-4.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "command-r7b:7b-12-2024-q4_K_M",
      "size_bytes": 5057009504,
      "model_version_id": 2,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 74.798,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.3469,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.94,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0507,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 3.3507,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 70.12,
      "ci_high": 79.11,
      "hard_score": 39.6578,
      "hard_n": 38,
      "ext_overall": 74.798,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 116,
      "class_rank": 34,
      "rank_lo": 87,
      "rank_hi": 152,
      "class_rank_lo": 23,
      "class_rank_hi": 52,
      "categories": {
        "reasoning": 71.4286,
        "math": 72.0,
        "coding": 59.3646,
        "instruction": 82.8889,
        "knowledge": 86.6667,
        "long_context": 83.3333,
        "multilingual": 96.6667,
        "tool_use": 25.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5936,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8289,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4667,
          "delta_mean": 1.1667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 0.8286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 1.0833,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.6634,
        "decode_tps_c4": 14.0239,
        "ttft_s": 0.557,
        "prefill_tps_4k": 265.3425,
        "prefill_tps_12k": 266.0527,
        "load_seconds": 60.2003,
        "size_gb": 10.4233,
        "vram_gb": 10.4233,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codestral-22b-v0-1",
      "display_name": "Codestral 22B V0.1",
      "family": "mistral",
      "org": null,
      "params_b": 22.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "codestral:22b-v0.1-q4_K_M",
      "size_bytes": 13341000000,
      "model_version_id": 200,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 74.7525,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.6864,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.3445,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0676,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 1.2614,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 69.7,
      "ci_high": 79.54,
      "hard_score": 39.6258,
      "hard_n": 35,
      "ext_overall": 74.7525,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 117,
      "class_rank": 45,
      "rank_lo": 80,
      "rank_hi": 154,
      "class_rank_lo": 41,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 45.7143,
        "math": 48.0,
        "coding": 72.9472,
        "instruction": 74.4444,
        "knowledge": 96.6667,
        "long_context": 83.3333,
        "multilingual": 83.3333,
        "tool_use": 85.0,
        "safety": 83.3333,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7295,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8333,
          "delta_mean": 1.1333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1,
          "delta_mean": 1.8,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9143,
          "delta_mean": 0.5714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.0833,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.4313,
        "decode_tps_c4": 6.5467,
        "ttft_s": 1.1655,
        "prefill_tps_4k": 76.1053,
        "prefill_tps_12k": 64.4064,
        "load_seconds": 241.9736,
        "size_gb": 43.9631,
        "vram_gb": 43.9631,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-meta-llama-3-8b-instruct",
      "display_name": "Meta-Llama-3-8B-Instruct",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 8.03,
      "arch": "dense",
      "hf_downloads": 927792,
      "hf_rank": 178,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama3:8b-instruct-q4_K_M",
      "size_bytes": 4920733984,
      "model_version_id": 20,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 74.7278,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.6817,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8436,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0876,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 5.8291,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 69.87,
      "ci_high": 79.46,
      "hard_score": 35.7022,
      "hard_n": 38,
      "ext_overall": 74.7278,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 118,
      "class_rank": 35,
      "rank_lo": 87,
      "rank_hi": 152,
      "class_rank_lo": 23,
      "class_rank_hi": 52,
      "categories": {
        "reasoning": 51.4286,
        "math": 36.0,
        "coding": 60.0107,
        "instruction": 89.2778,
        "knowledge": 90.0,
        "long_context": 83.3333,
        "multilingual": 81.6667,
        "tool_use": 85.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6001,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6667,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8928,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0333,
          "delta_mean": 1.0333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.36,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1429,
          "delta_mean": 0.9714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.75,
          "delta_mean": 1.2083,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.5772,
        "decode_tps_c4": 27.2717,
        "ttft_s": 0.4067,
        "prefill_tps_4k": 272.2075,
        "prefill_tps_12k": 271.2225,
        "load_seconds": 38.1772,
        "size_gb": 9.1557,
        "vram_gb": 9.1557,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepcoder-14b-preview",
      "display_name": "Deepcoder 14B Preview",
      "family": "other",
      "org": null,
      "params_b": 14.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "DeepSeek-R1-Distill-Qwen 14B",
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepcoder:14b-preview-q4_K_M",
      "size_bytes": 9000000000,
      "model_version_id": 83,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 74.4947,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.5686,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4526,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9117,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 1.9511,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 69.53,
      "ci_high": 79.02,
      "hard_score": 36.6901,
      "hard_n": 71,
      "ext_overall": 85.015,
      "n_truncated": 51,
      "n_extended": 51,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        72,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 119,
      "class_rank": 18,
      "rank_lo": 84,
      "rank_hi": 146,
      "class_rank_lo": 13,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 74.2857,
        "math": 56.0,
        "coding": 20.0,
        "instruction": 86.8333,
        "knowledge": 100.0,
        "long_context": 66.6667,
        "multilingual": 75.0,
        "tool_use": 100.0,
        "safety": 91.6667,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.2,
          "delta_mean": 0.2,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8683,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7667,
          "delta_mean": 1.1,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1714,
          "delta_mean": 0.8,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 1.4583,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.6966,
        "decode_tps_c4": 10.3271,
        "ttft_s": 0.6709,
        "prefill_tps_4k": 145.7093,
        "prefill_tps_12k": 107.9164,
        "load_seconds": 198.953,
        "size_gb": 34.8887,
        "vram_gb": 34.8887,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "google-medgemma-4b-it",
      "display_name": "medgemma-4b-it",
      "family": "gemma",
      "org": "google",
      "params_b": 4.3,
      "arch": "dense",
      "hf_downloads": 907335,
      "hf_rank": 184,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "medgemma:4b-it-q4_K_M",
      "size_bytes": 3338929952,
      "model_version_id": 13,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 74.3436,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.3733,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.7976,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1429,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 9.2414,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "medical",
      "specialised_label": "medical",
      "withheld": null,
      "ci_low": 69.33,
      "ci_high": 79.17,
      "hard_score": 38.5883,
      "hard_n": 38,
      "ext_overall": 74.4018,
      "n_truncated": 18,
      "n_extended": 18,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 120,
      "class_rank": 36,
      "rank_lo": 88,
      "rank_hi": 152,
      "class_rank_lo": 23,
      "class_rank_hi": 52,
      "categories": {
        "reasoning": 65.7143,
        "math": 72.0,
        "coding": 60.323,
        "instruction": 78.5556,
        "knowledge": 80.0,
        "long_context": 50.0,
        "multilingual": 77.5,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6032,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7856,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.5,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3714,
          "delta_mean": 0.7143,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.25,
          "delta_mean": 1.4583,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.103,
        "decode_tps_c4": 42.184,
        "ttft_s": 0.3097,
        "prefill_tps_4k": 490.1869,
        "prefill_tps_12k": 491.2058,
        "load_seconds": 26.5966,
        "size_gb": 4.4596,
        "vram_gb": 4.4596,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lfm2-5-thinking-1-2b",
      "display_name": "Lfm2.5 Thinking 1.2B",
      "family": "lfm",
      "org": null,
      "params_b": 1.2,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "lfm2.5-thinking:1.2b-q4_K_M",
      "size_bytes": 731000000,
      "model_version_id": 132,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 74.3049,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.3183,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.0938,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8783,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 63.7634,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 69.46,
      "ci_high": 79.05,
      "hard_score": 45.2106,
      "hard_n": 71,
      "ext_overall": 78.0935,
      "n_truncated": 32,
      "n_extended": 32,
      "n_gated": 2,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 121,
      "class_rank": 5,
      "rank_lo": 84,
      "rank_hi": 148,
      "class_rank_lo": 1,
      "class_rank_hi": 9,
      "categories": {
        "reasoning": 82.8571,
        "math": 80.0,
        "coding": 42.0539,
        "instruction": 80.7778,
        "knowledge": 83.3333,
        "long_context": 38.8889,
        "multilingual": 71.6667,
        "tool_use": 97.5,
        "safety": 91.6667,
        "sycophancy": 50.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4205,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6667,
          "delta_mean": 0.4,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8078,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6667,
          "delta_mean": 1.0667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4688,
          "delta_mean": 0.55,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.975,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.75,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 204.953,
        "decode_tps_c4": 122.0261,
        "ttft_s": 0.0296,
        "prefill_tps_4k": 6054.8439,
        "prefill_tps_12k": 6498.2727,
        "load_seconds": 3.0383,
        "size_gb": 1.0978,
        "vram_gb": 1.0978,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-3b",
      "display_name": "Granite4 3B",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite4:3b",
      "size_bytes": 2100000000,
      "model_version_id": 125,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 74.1381,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.6283,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 6.1862,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9683,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 20.0636,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 68.98,
      "ci_high": 79.21,
      "hard_score": 41.3955,
      "hard_n": 71,
      "ext_overall": 74.1381,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 122,
      "class_rank": 10,
      "rank_lo": 85,
      "rank_hi": 145,
      "class_rank_lo": 3,
      "class_rank_hi": 17,
      "categories": {
        "reasoning": 62.8571,
        "math": 56.0,
        "coding": 64.6637,
        "instruction": 87.3333,
        "knowledge": 76.6667,
        "long_context": 55.5556,
        "multilingual": 78.3333,
        "tool_use": 90.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6466,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8733,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2333,
          "delta_mean": 0.9,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1143,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0833,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 71.2668,
        "decode_tps_c4": 177.132,
        "ttft_s": 0.0602,
        "prefill_tps_4k": 1586.1278,
        "prefill_tps_12k": 1229.4174,
        "load_seconds": 48.938,
        "size_gb": 13.1857,
        "vram_gb": 13.1857,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-2-8b-instruct",
      "display_name": "Granite3.2 8B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.2:8b-instruct-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 115,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 74.1207,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.3238,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8869,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0512,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 6.8246,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 69.49,
      "ci_high": 78.94,
      "hard_score": 31.0256,
      "hard_n": 71,
      "ext_overall": 74.1207,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        102,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        116,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        136,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        160,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 123,
      "class_rank": 37,
      "rank_lo": 85,
      "rank_hi": 145,
      "class_rank_lo": 22,
      "class_rank_hi": 48,
      "categories": {
        "reasoning": 45.7143,
        "math": 56.0,
        "coding": 58.5385,
        "instruction": 82.6667,
        "knowledge": 96.6667,
        "long_context": 50.0,
        "multilingual": 81.6667,
        "tool_use": 100.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5854,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8267,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 1.3,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9429,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0417,
          "delta_mean": 1.3333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 34.6716,
        "decode_tps_c4": 97.6726,
        "ttft_s": 0.1334,
        "prefill_tps_4k": 835.7712,
        "prefill_tps_12k": 714.0253,
        "load_seconds": 100.438,
        "size_gb": 26.9319,
        "vram_gb": 26.9319,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-ai-deepseek-r1-distill-qwen-7b",
      "display_name": "DeepSeek-R1-Distill-Qwen-7B",
      "family": "deepseek",
      "org": "deepseek-ai",
      "params_b": 7.62,
      "arch": "dense",
      "hf_downloads": 346084,
      "hf_rank": 246,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:7b-qwen-distill-q4_K_M",
      "size_bytes": 4683073184,
      "model_version_id": 233,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 73.6455,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.1305,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7064,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8131,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 4.5339,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 68.92,
      "ci_high": 78.3,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 77.6697,
      "n_truncated": 27,
      "n_extended": 27,
      "n_gated": 2,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        195,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 124,
      "class_rank": 38,
      "rank_lo": 82,
      "rank_hi": 155,
      "class_rank_lo": 20,
      "class_rank_hi": 53,
      "categories": {
        "reasoning": 77.1429,
        "math": 96.0,
        "coding": 37.0,
        "instruction": 77.3889,
        "knowledge": 83.3333,
        "long_context": 44.4444,
        "multilingual": 80.8333,
        "tool_use": 100.0,
        "safety": 66.6667,
        "sycophancy": 75.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.37,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4,
          "delta_mean": 0.2667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7739,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9333,
          "delta_mean": 1.2667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6571,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0417,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.8388,
        "decode_tps_c4": 20.6627,
        "ttft_s": 0.3101,
        "prefill_tps_4k": 341.5158,
        "prefill_tps_12k": 301.1161,
        "load_seconds": 54.1717,
        "size_gb": 4.6831,
        "vram_gb": 4.6831,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "glm4-9b-chat",
      "display_name": "Glm4 9B Chat",
      "family": "glm",
      "org": null,
      "params_b": 9.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "glm4:9b-chat-q4_K_M",
      "size_bytes": 6300000000,
      "model_version_id": 107,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 73.3139,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 6.3819,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8421,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9826,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 3.4874,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 68.01,
      "ci_high": 78.53,
      "hard_score": 38.3279,
      "hard_n": 71,
      "ext_overall": 73.3139,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        123,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 125,
      "class_rank": 39,
      "rank_lo": 87,
      "rank_hi": 153,
      "class_rank_lo": 24,
      "class_rank_hi": 52,
      "categories": {
        "reasoning": 60.0,
        "math": 60.0,
        "coding": 59.9922,
        "instruction": 82.3333,
        "knowledge": 90.0,
        "long_context": 66.6667,
        "multilingual": 85.0,
        "tool_use": 60.0,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5999,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8333,
          "delta_mean": 1.0333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8233,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5333,
          "delta_mean": 1.2333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8857,
          "delta_mean": 0.7714,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4583,
          "delta_mean": 1.0,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.8317,
        "decode_tps_c4": 15.9224,
        "ttft_s": 0.4128,
        "prefill_tps_4k": 261.2781,
        "prefill_tps_12k": 228.4236,
        "load_seconds": 40.578,
        "size_gb": 11.8024,
        "vram_gb": 11.8024,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "tulu3-8b",
      "display_name": "Tulu3 8B",
      "family": "other",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Llama 3.1",
      "license": "llama3.1",
      "quantization": "Q4_K_M",
      "ollama_ref": "tulu3:8b-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 183,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 72.9871,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 6.4486,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8755,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1136,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 3.454,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.94,
      "ci_high": 77.91,
      "hard_score": 37.7544,
      "hard_n": 71,
      "ext_overall": 72.9871,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 126,
      "class_rank": 40,
      "rank_lo": 87,
      "rank_hi": 152,
      "class_rank_lo": 24,
      "class_rank_hi": 51,
      "categories": {
        "reasoning": 60.0,
        "math": 52.0,
        "coding": 52.7176,
        "instruction": 83.5556,
        "knowledge": 90.0,
        "long_context": 61.1111,
        "multilingual": 67.5,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5272,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1667,
          "delta_mean": 1.2,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8356,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2667,
          "delta_mean": 1.3,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.675,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4857,
          "delta_mean": 0.9429,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0833,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.0478,
        "decode_tps_c4": 28.2358,
        "ttft_s": 0.3732,
        "prefill_tps_4k": 273.9014,
        "prefill_tps_12k": 231.1394,
        "load_seconds": 126.469,
        "size_gb": 22.351,
        "vram_gb": 22.351,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "dphn-dolphin-2-9-1-yi-1-5-34b",
      "display_name": "dolphin-2.9.1-yi-1.5-34b",
      "family": "other",
      "org": "dphn",
      "params_b": 34.39,
      "arch": "dense",
      "hf_downloads": 4725752,
      "hf_rank": 29,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/bartowski/dolphin-2.9.1-yi-1.5-34b-GGUF:Q4_K_M",
      "size_bytes": null,
      "model_version_id": 4,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 72.964,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.1602,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.8034,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1413,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 0.8213,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.86,
      "ci_high": 78.2,
      "hard_score": 31.7025,
      "hard_n": 38,
      "ext_overall": 72.964,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        21,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 127,
      "class_rank": 46,
      "rank_lo": 90,
      "rank_hi": 154,
      "class_rank_lo": 42,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 71.4286,
        "math": 56.0,
        "coding": 65.6918,
        "instruction": 72.7222,
        "knowledge": 93.3333,
        "long_context": 66.6667,
        "multilingual": 66.6667,
        "tool_use": 85.0,
        "safety": 79.1667,
        "sycophancy": 91.6667,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6569,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0667,
          "delta_mean": 1.3667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7272,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5333,
          "delta_mean": 1.3333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8571,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4348,
          "delta_mean": 0.9565,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 2.4005,
        "decode_tps_c4": 7.191,
        "ttft_s": 1.9447,
        "prefill_tps_4k": 50.7374,
        "prefill_tps_12k": 51.347,
        "load_seconds": 123.0628,
        "size_gb": 37.2184,
        "vram_gb": 37.2184,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-1-dense-8b-instruct",
      "display_name": "Granite3.1 Dense 8B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.1-dense:8b-instruct-q4_K_M",
      "size_bytes": 5000000000,
      "model_version_id": 111,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 72.8603,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 6.4517,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.9276,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.095,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 3.9912,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.81,
      "ci_high": 77.64,
      "hard_score": 31.7793,
      "hard_n": 71,
      "ext_overall": 72.8603,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        93,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        119,
        120,
        121,
        123,
        124,
        128,
        130,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 128,
      "class_rank": 41,
      "rank_lo": 88,
      "rank_hi": 151,
      "class_rank_lo": 25,
      "class_rank_hi": 50,
      "categories": {
        "reasoning": 57.1429,
        "math": 52.0,
        "coding": 64.211,
        "instruction": 74.3333,
        "knowledge": 100.0,
        "long_context": 55.5556,
        "multilingual": 61.6667,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 50.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6421,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7433,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.4,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5714,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8333,
          "delta_mean": 1.5,
          "disputed": 5,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 34.8098,
        "decode_tps_c4": 97.2367,
        "ttft_s": 0.1288,
        "prefill_tps_4k": 817.7488,
        "prefill_tps_12k": 685.1801,
        "load_seconds": 105.094,
        "size_gb": 26.9807,
        "vram_gb": 26.9807,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "aya-expanse-8b",
      "display_name": "Aya Expanse 8B",
      "family": "cohere",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "cc-by-nc-4.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "aya-expanse:8b-q4_K_M",
      "size_bytes": 5100000000,
      "model_version_id": 71,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 72.5476,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.9821,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.694,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.059,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 3.2435,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.54,
      "ci_high": 77.26,
      "hard_score": 35.4914,
      "hard_n": 71,
      "ext_overall": 72.5476,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        123,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 129,
      "class_rank": 42,
      "rank_lo": 90,
      "rank_hi": 154,
      "class_rank_lo": 27,
      "class_rank_hi": 53,
      "categories": {
        "reasoning": 54.2857,
        "math": 40.0,
        "coding": 49.9762,
        "instruction": 79.5,
        "knowledge": 90.0,
        "long_context": 66.6667,
        "multilingual": 72.5,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4998,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9667,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.795,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7,
          "delta_mean": 1.5333,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.725,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4286,
          "delta_mean": 0.4286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.0,
          "delta_mean": 1.4583,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.6996,
        "decode_tps_c4": 16.2824,
        "ttft_s": 0.4011,
        "prefill_tps_4k": 243.9431,
        "prefill_tps_12k": 243.1439,
        "load_seconds": 82.187,
        "size_gb": 14.1562,
        "vram_gb": 14.1562,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-coder-3b-instruct",
      "display_name": "Qwen2.5-Coder-3B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 3.09,
      "arch": "dense",
      "hf_downloads": 530170,
      "hf_rank": 297,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "qwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:3b-instruct-q4_K_M",
      "size_bytes": 1929903072,
      "model_version_id": 38,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 72.5305,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.6638,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9538,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1733,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 12.0131,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 67.71,
      "ci_high": 77.46,
      "hard_score": 29.9021,
      "hard_n": 38,
      "ext_overall": 72.5305,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        29,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 130,
      "class_rank": 11,
      "rank_lo": 89,
      "rank_hi": 155,
      "class_rank_lo": 4,
      "class_rank_hi": 17,
      "categories": {
        "reasoning": 57.1429,
        "math": 44.0,
        "coding": 60.2425,
        "instruction": 70.0,
        "knowledge": 80.0,
        "long_context": 72.2222,
        "multilingual": 78.3333,
        "tool_use": 95.0,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6024,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5667,
          "delta_mean": 1.0667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.6333,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.44,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6857,
          "delta_mean": 1.0,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.75,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 21.3091,
        "decode_tps_c4": 49.332,
        "ttft_s": 0.2076,
        "prefill_tps_4k": 665.1641,
        "prefill_tps_12k": 579.0787,
        "load_seconds": 22.7514,
        "size_gb": 4.7279,
        "vram_gb": 4.7279,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-5-vl-3b-instruct",
      "display_name": "Qwen2.5-VL-3B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 3.75,
      "arch": "dense",
      "hf_downloads": 2292740,
      "hf_rank": 65,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5vl:3b-q4_K_M",
      "size_bytes": 3200614720,
      "model_version_id": 41,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 72.3435,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.7457,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.031,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0552,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 11.9909,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.22,
      "ci_high": 77.59,
      "hard_score": 22.2028,
      "hard_n": 38,
      "ext_overall": 72.5699,
      "n_truncated": 8,
      "n_extended": 8,
      "n_gated": 1,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 131,
      "class_rank": 12,
      "rank_lo": 90,
      "rank_hi": 157,
      "class_rank_lo": 5,
      "class_rank_hi": 18,
      "categories": {
        "reasoning": 71.4286,
        "math": 64.0,
        "coding": 42.6076,
        "instruction": 74.7222,
        "knowledge": 100.0,
        "long_context": 66.6667,
        "multilingual": 60.0,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4261,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5667,
          "delta_mean": 0.8667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7472,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 1.4667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3714,
          "delta_mean": 0.9429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 21.9232,
        "decode_tps_c4": 51.4279,
        "ttft_s": 0.2082,
        "prefill_tps_4k": 658.9922,
        "prefill_tps_12k": 574.4994,
        "load_seconds": 24.6148,
        "size_gb": 5.4668,
        "vram_gb": 5.4668,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-qwen2-vl-7b-instruct",
      "display_name": "Qwen2-VL-7B-Instruct",
      "family": "qwen",
      "org": "Qwen",
      "params_b": 8.29,
      "arch": "dense",
      "hf_downloads": 610590,
      "hf_rank": 258,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/bartowski/Qwen2-VL-7B-Instruct-GGUF:Q4_K_M",
      "size_bytes": null,
      "model_version_id": 43,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 72.242,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.5086,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9826,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9474,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.8837,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.39,
      "ci_high": 76.99,
      "hard_score": 21.6182,
      "hard_n": 38,
      "ext_overall": 72.242,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 0,
      "beats": [
        17,
        22,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 132,
      "class_rank": 43,
      "rank_lo": 92,
      "rank_hi": 157,
      "class_rank_lo": 25,
      "class_rank_hi": 54,
      "categories": {
        "reasoning": 45.7143,
        "math": 64.0,
        "coding": 49.7416,
        "instruction": 70.7222,
        "knowledge": 86.6667,
        "long_context": 66.6667,
        "multilingual": 66.6667,
        "tool_use": 100.0,
        "safety": 100.0,
        "sycophancy": 66.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4974,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1333,
          "delta_mean": 1.0333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7072,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2333,
          "delta_mean": 1.3667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1714,
          "delta_mean": 0.6286,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.875,
          "delta_mean": 0.9583,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.1797,
        "decode_tps_c4": 16.1233,
        "ttft_s": 0.4258,
        "prefill_tps_4k": 331.1762,
        "prefill_tps_12k": 294.2369,
        "load_seconds": 55.3998,
        "size_gb": 12.7876,
        "vram_gb": 12.7876,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lfm3-vl-2b",
      "display_name": "Lfm3 VL 2B",
      "family": "lfm",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "lfm3-vl:2b-q4_K_M",
      "size_bytes": 1500000000,
      "model_version_id": 216,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 72.2366,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 6.1467,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.5112,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9517,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 11.0317,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 67.22,
      "ci_high": 77.03,
      "hard_score": 20.8817,
      "hard_n": 35,
      "ext_overall": 72.3189,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 5,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 133,
      "class_rank": 6,
      "rank_lo": 89,
      "rank_hi": 158,
      "class_rank_lo": 1,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 62.8571,
        "math": 60.0,
        "coding": 42.2165,
        "instruction": 88.1111,
        "knowledge": 86.6667,
        "long_context": 44.4444,
        "multilingual": 91.6667,
        "tool_use": 95.0,
        "safety": 79.1667,
        "sycophancy": 100.0,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7,
          "delta_mean": 0.8333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8811,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0333,
          "delta_mean": 1.2,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1143,
          "delta_mean": 0.6,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8333,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 41.8658,
        "decode_tps_c4": 58.6535,
        "ttft_s": 0.1503,
        "prefill_tps_4k": 1285.0103,
        "prefill_tps_12k": 1259.0864,
        "load_seconds": 17.3012,
        "size_gb": 2.0083,
        "vram_gb": 2.0083,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "medgemma1-5-4b-it",
      "display_name": "Medgemma1.5 4B IT",
      "family": "gemma",
      "org": null,
      "params_b": 4.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": "Gemma 3",
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "medgemma1.5:4b-it-q4_K_M",
      "size_bytes": 3300000000,
      "model_version_id": 137,
      "run_id": "r20261029-bv1-3e4075",
      "battery_version": "v1",
      "tested_at": "2026-10-29T14:57:58.390Z",
      "overall": 71.5704,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 5.4662,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.0662,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.021,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 5.8361,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": "medical",
      "specialised_label": "medical",
      "withheld": null,
      "ci_low": 66.62,
      "ci_high": 76.42,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 72.7404,
      "n_truncated": 31,
      "n_extended": 28,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 134,
      "class_rank": 13,
      "rank_lo": 91,
      "rank_hi": 159,
      "class_rank_lo": 4,
      "class_rank_hi": 20,
      "categories": {
        "reasoning": 65.7143,
        "math": 48.0,
        "coding": 56.8639,
        "instruction": 52.1667,
        "knowledge": 86.6667,
        "long_context": 72.2222,
        "multilingual": 74.1667,
        "tool_use": 92.5,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5686,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4333,
          "delta_mean": 0.6333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5217,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5333,
          "delta_mean": 1.5333,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1143,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.5,
          "delta_mean": 1.2917,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 20.1878,
        "decode_tps_c4": 30.6431,
        "ttft_s": 0.2759,
        "prefill_tps_4k": 508.4888,
        "prefill_tps_12k": 507.9796,
        "load_seconds": 39.3504,
        "size_gb": 5.9444,
        "vram_gb": 5.9444,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "phi4-mini-reasoning-3-8b",
      "display_name": "Phi4 Mini Reasoning 3.8B",
      "family": "phi",
      "org": null,
      "params_b": 3.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi4-mini-reasoning:3.8b-q4_K_M",
      "size_bytes": 3200000000,
      "model_version_id": 151,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 70.9406,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 4.4487,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8967,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2102,
      "agreement": "high",
      "disputed": 20,
      "tokens_per_s": 17.4302,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 65.92,
      "ci_high": 76.0,
      "hard_score": 47.8809,
      "hard_n": 71,
      "ext_overall": 74.4009,
      "n_truncated": 37,
      "n_extended": 37,
      "n_gated": 1,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 135,
      "class_rank": 14,
      "rank_lo": 97,
      "rank_hi": 157,
      "class_rank_lo": 4,
      "class_rank_hi": 18,
      "categories": {
        "reasoning": 94.2857,
        "math": 96.0,
        "coding": 31.8182,
        "instruction": 44.2778,
        "knowledge": 90.0,
        "long_context": 66.6667,
        "multilingual": 75.0,
        "tool_use": 55.0,
        "safety": 85.4167,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3182,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.9,
          "delta_mean": 1.2,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4428,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5862,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.7,
          "delta_mean": 2.0345,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.96,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.9429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.8,
          "delta_mean": 0.4,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8542,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5833,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.55,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 48.8338,
        "decode_tps_c4": 139.1304,
        "ttft_s": 0.0665,
        "prefill_tps_4k": 1945.7809,
        "prefill_tps_12k": 1410.7968,
        "load_seconds": 78.016,
        "size_gb": 20.6932,
        "vram_gb": 20.6932,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "yi-9b-chat-v1-5",
      "display_name": "YI 9B Chat V1.5",
      "family": "yi",
      "org": null,
      "params_b": 9.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "yi:9b-chat-v1.5-q4_K_M",
      "size_bytes": 5300000000,
      "model_version_id": 185,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 70.474,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.8307,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4502,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1505,
      "agreement": "high",
      "disputed": 20,
      "tokens_per_s": 3.6878,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 65.41,
      "ci_high": 75.29,
      "hard_score": 37.9224,
      "hard_n": 71,
      "ext_overall": 70.474,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        144,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 136,
      "class_rank": 44,
      "rank_lo": 103,
      "rank_hi": 156,
      "class_rank_lo": 30,
      "class_rank_hi": 53,
      "categories": {
        "reasoning": 68.5714,
        "math": 64.0,
        "coding": 57.472,
        "instruction": 73.6667,
        "knowledge": 90.0,
        "long_context": 22.2222,
        "multilingual": 75.0,
        "tool_use": 87.5,
        "safety": 95.8333,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5747,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3,
          "delta_mean": 0.9333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7367,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2333,
          "delta_mean": 1.6333,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3429,
          "delta_mean": 0.6857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.875,
          "delta_mean": 1.375,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.9561,
        "decode_tps_c4": 24.5729,
        "ttft_s": 0.5477,
        "prefill_tps_4k": 229.0683,
        "prefill_tps_12k": 242.0483,
        "load_seconds": 33.0,
        "size_gb": 8.612,
        "vram_gb": 8.612,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "marco-o1-7b",
      "display_name": "Marco O1 7B",
      "family": "other",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2 7B",
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "marco-o1:7b-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 135,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 70.4729,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 5.546,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.8606,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2197,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 4.6737,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 65.05,
      "ci_high": 76.04,
      "hard_score": 37.2982,
      "hard_n": 71,
      "ext_overall": 70.4729,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 137,
      "class_rank": 45,
      "rank_lo": 97,
      "rank_hi": 157,
      "class_rank_lo": 29,
      "class_rank_hi": 54,
      "categories": {
        "reasoning": 71.4286,
        "math": 64.0,
        "coding": 64.7168,
        "instruction": 46.8889,
        "knowledge": 76.6667,
        "long_context": 72.2222,
        "multilingual": 69.1667,
        "tool_use": 90.0,
        "safety": 79.1667,
        "sycophancy": 91.6667,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6472,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1333,
          "delta_mean": 0.9,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4689,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6333,
          "delta_mean": 1.3,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8571,
          "delta_mean": 1.1429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3043,
          "delta_mean": 1.1304,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 2,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.375,
          "delta_mean": 1.625,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.8004,
        "decode_tps_c4": 23.7108,
        "ttft_s": 0.3707,
        "prefill_tps_4k": 315.7247,
        "prefill_tps_12k": 275.6601,
        "load_seconds": 58.39,
        "size_gb": 12.5369,
        "vram_gb": 12.5369,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "cogito-3b-v1-preview-llama",
      "display_name": "Cogito 3B V1 Preview Llama",
      "family": "llama",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Llama 3.2",
      "license": "llama3.2",
      "quantization": "Q4_K_M",
      "ollama_ref": "cogito:3b-v1-preview-llama-q4_K_M",
      "size_bytes": 2200000000,
      "model_version_id": 79,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 70.3849,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.9557,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4776,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9295,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 19.8763,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 65.38,
      "ci_high": 75.25,
      "hard_score": 32.6068,
      "hard_n": 71,
      "ext_overall": 70.3849,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        114,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 138,
      "class_rank": 15,
      "rank_lo": 104,
      "rank_hi": 157,
      "class_rank_lo": 6,
      "class_rank_hi": 18,
      "categories": {
        "reasoning": 45.7143,
        "math": 40.0,
        "coding": 49.2502,
        "instruction": 81.5556,
        "knowledge": 86.6667,
        "long_context": 61.1111,
        "multilingual": 76.6667,
        "tool_use": 92.5,
        "safety": 100.0,
        "sycophancy": 66.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9,
          "delta_mean": 1.0,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8156,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9333,
          "delta_mean": 1.4667,
          "disputed": 10,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9714,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3333,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.375,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 73.3802,
        "decode_tps_c4": 157.9272,
        "ttft_s": 0.0539,
        "prefill_tps_4k": 2152.74,
        "prefill_tps_12k": 1600.5926,
        "load_seconds": 68.829,
        "size_gb": 17.4131,
        "vram_gb": 17.4131,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "lfm3-1-7b",
      "display_name": "Lfm3 1.7B",
      "family": "lfm",
      "org": null,
      "params_b": 1.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "lfm3:1.7b-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 215,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 70.2464,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 6.1193,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.7826,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9376,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 11.4907,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 65.33,
      "ci_high": 75.33,
      "hard_score": 29.6863,
      "hard_n": 35,
      "ext_overall": 76.9517,
      "n_truncated": 45,
      "n_extended": 45,
      "n_gated": 1,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        199,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 139,
      "class_rank": 7,
      "rank_lo": 96,
      "rank_hi": 162,
      "class_rank_lo": 1,
      "class_rank_hi": 14,
      "categories": {
        "reasoning": 74.2857,
        "math": 76.0,
        "coding": 61.5431,
        "instruction": 77.8889,
        "knowledge": 90.0,
        "long_context": 33.3333,
        "multilingual": 83.3333,
        "tool_use": 40.0,
        "safety": 95.8333,
        "sycophancy": 100.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6154,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1667,
          "delta_mean": 0.7667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7789,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5667,
          "delta_mean": 1.4667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9714,
          "delta_mean": 0.3714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2083,
          "delta_mean": 1.2083,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 42.1177,
        "decode_tps_c4": 58.1799,
        "ttft_s": 0.1509,
        "prefill_tps_4k": 1275.5013,
        "prefill_tps_12k": 1267.8843,
        "load_seconds": 18.957,
        "size_gb": 1.6181,
        "vram_gb": 1.6181,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "olmo2-13b-1124-instruct",
      "display_name": "Olmo2 13B 1124 Instruct",
      "family": "olmo",
      "org": null,
      "params_b": 13.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "olmo2:13b-1124-instruct-q4_K_M",
      "size_bytes": 8400000000,
      "model_version_id": 143,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 69.5903,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 6.5264,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.926,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0019,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 3.4266,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 64.54,
      "ci_high": 74.51,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 69.5903,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 140,
      "class_rank": 19,
      "rank_lo": 97,
      "rank_hi": 165,
      "class_rank_lo": 15,
      "class_rank_hi": 19,
      "categories": {
        "reasoning": 60.0,
        "math": 52.0,
        "coding": 40.7572,
        "instruction": 88.5556,
        "knowledge": 96.6667,
        "long_context": 33.3333,
        "multilingual": 80.0,
        "tool_use": 75.0,
        "safety": 100.0,
        "sycophancy": 100.0,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4076,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1333,
          "delta_mean": 1.3333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8856,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7333,
          "delta_mean": 1.3,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.9714,
          "delta_mean": 0.5429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.5417,
          "delta_mean": 1.0833,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.8386,
        "decode_tps_c4": 15.4968,
        "ttft_s": 0.6257,
        "prefill_tps_4k": 149.2482,
        "prefill_tps_12k": 161.2257,
        "load_seconds": 109.453,
        "size_gb": 21.7457,
        "vram_gb": 21.7457,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codegeex4-9b-all",
      "display_name": "Codegeex4 9B All",
      "family": "other",
      "org": null,
      "params_b": 9.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "codegeex4:9b-all-q4_K_M",
      "size_bytes": 6300000000,
      "model_version_id": 72,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 68.678,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 5.5002,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.871,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0236,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 3.3565,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 63.18,
      "ci_high": 73.79,
      "hard_score": 39.5436,
      "hard_n": 71,
      "ext_overall": 68.678,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 2,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 141,
      "class_rank": 46,
      "rank_lo": 109,
      "rank_hi": 159,
      "class_rank_lo": 32,
      "class_rank_hi": 55,
      "categories": {
        "reasoning": 48.5714,
        "math": 36.0,
        "coding": 60.531,
        "instruction": 75.2222,
        "knowledge": 93.3333,
        "long_context": 77.7778,
        "multilingual": 63.3333,
        "tool_use": 80.0,
        "safety": 83.3333,
        "sycophancy": 83.3333,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6053,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7522,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2333,
          "delta_mean": 1.5667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.36,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7714,
          "delta_mean": 0.5429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.0417,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.375,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.0521,
        "decode_tps_c4": 16.2661,
        "ttft_s": 0.4071,
        "prefill_tps_4k": 261.6387,
        "prefill_tps_12k": 229.017,
        "load_seconds": 40.281,
        "size_gb": 11.8024,
        "vram_gb": 11.8024,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-3-2b",
      "display_name": "Granite3.3 2B",
      "family": "granite",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.3:2b",
      "size_bytes": 1500000000,
      "model_version_id": 116,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 68.642,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 6.011,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.5671,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.99,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 7.9953,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 63.8,
      "ci_high": 73.81,
      "hard_score": 30.3826,
      "hard_n": 71,
      "ext_overall": 68.4568,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        150,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 142,
      "class_rank": 8,
      "rank_lo": 111,
      "rank_hi": 159,
      "class_rank_lo": 6,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 57.1429,
        "math": 52.0,
        "coding": 50.3575,
        "instruction": 76.3333,
        "knowledge": 90.0,
        "long_context": 44.4444,
        "multilingual": 57.5,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5036,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7667,
          "delta_mean": 1.0333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7633,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4333,
          "delta_mean": 1.5667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.575,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8857,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.375,
          "delta_mean": 1.375,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 0.375,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 28.5526,
        "decode_tps_c4": 58.5143,
        "ttft_s": 0.1864,
        "prefill_tps_4k": 645.0414,
        "prefill_tps_12k": 546.4601,
        "load_seconds": 77.968,
        "size_gb": 12.6166,
        "vram_gb": 12.6166,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-5-1-5b-instruct",
      "display_name": "Qwen2.5 1.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:1.5b-instruct-q4_K_M",
      "size_bytes": 986000000,
      "model_version_id": 160,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 68.5148,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.3781,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.6255,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.216,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 44.7851,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 63.16,
      "ci_high": 73.66,
      "hard_score": 35.9831,
      "hard_n": 71,
      "ext_overall": 68.5148,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 2,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 143,
      "class_rank": 9,
      "rank_lo": 111,
      "rank_hi": 160,
      "class_rank_lo": 5,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 62.8571,
        "math": 52.0,
        "coding": 49.8872,
        "instruction": 59.3889,
        "knowledge": 93.3333,
        "long_context": 66.6667,
        "multilingual": 48.3333,
        "tool_use": 92.5,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4989,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5939,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9667,
          "delta_mean": 1.5667,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6857,
          "delta_mean": 0.9714,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 45.4934,
        "decode_tps_c4": 43.8323,
        "ttft_s": 0.1,
        "prefill_tps_4k": 1222.1021,
        "prefill_tps_12k": 1003.4702,
        "load_seconds": 27.036,
        "size_gb": 5.0759,
        "vram_gb": 5.0759,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma2-2b-instruct",
      "display_name": "Gemma2 2B Instruct",
      "family": "gemma",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma2:2b-instruct-q4_K_M",
      "size_bytes": 1700000000,
      "model_version_id": 102,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 68.1774,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 5.8521,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4652,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1031,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 7.2721,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 62.89,
      "ci_high": 73.32,
      "hard_score": 32.6364,
      "hard_n": 71,
      "ext_overall": 68.1774,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 144,
      "class_rank": 10,
      "rank_lo": 112,
      "rank_hi": 160,
      "class_rank_lo": 5,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 42.8571,
        "math": 44.0,
        "coding": 42.406,
        "instruction": 84.6111,
        "knowledge": 86.6667,
        "long_context": 55.5556,
        "multilingual": 80.0,
        "tool_use": 77.5,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4241,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6333,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8461,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6,
          "delta_mean": 1.2333,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.44,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3429,
          "delta_mean": 0.8571,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.625,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.5,
          "disputed": 0,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 27.5628,
        "decode_tps_c4": 39.8645,
        "ttft_s": 0.1922,
        "prefill_tps_4k": 718.6028,
        "prefill_tps_12k": 716.1897,
        "load_seconds": 45.859,
        "size_gb": 4.6806,
        "vram_gb": 4.6806,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mathstral-7b-v0-1",
      "display_name": "Mathstral 7B V0.1",
      "family": "other",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Mistral 7B",
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "mathstral:7b-v0.1-q4_K_M",
      "size_bytes": 4400000000,
      "model_version_id": 136,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 67.987,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 4.9119,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.3003,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1931,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 3.2712,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": null,
      "ci_low": 62.44,
      "ci_high": 73.52,
      "hard_score": 36.7797,
      "hard_n": 71,
      "ext_overall": 67.987,
      "n_truncated": 4,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        173,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 145,
      "class_rank": 47,
      "rank_lo": 112,
      "rank_hi": 160,
      "class_rank_lo": 33,
      "class_rank_hi": 55,
      "categories": {
        "reasoning": 65.7143,
        "math": 68.0,
        "coding": 56.5856,
        "instruction": 52.2778,
        "knowledge": 86.6667,
        "long_context": 55.5556,
        "multilingual": 80.0,
        "tool_use": 70.0,
        "safety": 77.0833,
        "sycophancy": 83.3333,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5659,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7,
          "delta_mean": 1.2667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5228,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9667,
          "delta_mean": 1.5333,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2857,
          "delta_mean": 0.7143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7708,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1739,
          "delta_mean": 0.8261,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.375,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.0406,
        "decode_tps_c4": 16.7825,
        "ttft_s": 0.4174,
        "prefill_tps_4k": 271.6139,
        "prefill_tps_12k": 223.6423,
        "load_seconds": 126.843,
        "size_gb": 22.0301,
        "vram_gb": 22.0301,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nousresearch-hermes-3-llama-3-1-8b",
      "display_name": "Hermes-3-Llama-3.1-8B",
      "family": "llama",
      "org": "NousResearch",
      "params_b": 8.03,
      "arch": "dense",
      "hf_downloads": 477636,
      "hf_rank": 325,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3",
      "quantization": "Q4_K_M",
      "ollama_ref": "hermes3:8b-llama3.1-q4_K_M",
      "size_bytes": 4920739584,
      "model_version_id": 29,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 67.6126,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 5.8674,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.5229,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0379,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 3.8277,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 8,
      "concurrency_text": "4 to 8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 62.37,
      "ci_high": 73.13,
      "hard_score": 16.8507,
      "hard_n": 38,
      "ext_overall": 67.983,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 146,
      "class_rank": 48,
      "rank_lo": 109,
      "rank_hi": 173,
      "class_rank_lo": 33,
      "class_rank_hi": 60,
      "categories": {
        "reasoning": 54.2857,
        "math": 52.0,
        "coding": 42.3392,
        "instruction": 73.7778,
        "knowledge": 93.3333,
        "long_context": 44.4444,
        "multilingual": 71.6667,
        "tool_use": 85.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4234,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1333,
          "delta_mean": 1.1333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7378,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6333,
          "delta_mean": 1.1667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5143,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.2083,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.5839,
        "decode_tps_c4": 21.2588,
        "ttft_s": 0.4468,
        "prefill_tps_4k": 275.7762,
        "prefill_tps_12k": 229.4447,
        "load_seconds": 72.9935,
        "size_gb": 13.7275,
        "vram_gb": 13.7275,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-dense-8b-instruct",
      "display_name": "Granite3 Dense 8B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3-dense:8b-instruct-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 119,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 67.564,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 5.9629,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4683,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1174,
      "agreement": "high",
      "disputed": 19,
      "tokens_per_s": 11.8436,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 62.75,
      "ci_high": 72.31,
      "hard_score": 29.0193,
      "hard_n": 71,
      "ext_overall": 67.8232,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 147,
      "class_rank": 49,
      "rank_lo": 117,
      "rank_hi": 161,
      "class_rank_lo": 36,
      "class_rank_hi": 55,
      "categories": {
        "reasoning": 54.2857,
        "math": 44.0,
        "coding": 59.1791,
        "instruction": 76.4444,
        "knowledge": 93.3333,
        "long_context": 33.3333,
        "multilingual": 52.5,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5918,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2333,
          "delta_mean": 0.9333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7644,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1667,
          "delta_mean": 1.7,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.44,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.525,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4,
          "delta_mean": 0.8286,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.9167,
          "delta_mean": 1.25,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.4557,
        "decode_tps_c4": 14.4574,
        "ttft_s": 0.4866,
        "prefill_tps_4k": 255.2237,
        "prefill_tps_12k": 256.7021,
        "load_seconds": 46.64,
        "size_gb": 10.5426,
        "vram_gb": 10.5426,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "internlm2-7b-chat-v2-5",
      "display_name": "Internlm2 7B Chat V2.5",
      "family": "internlm",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "internlm2:7b-chat-v2.5-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 130,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 67.3585,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 5.6814,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.4721,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2188,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 3.7116,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 62.06,
      "ci_high": 72.58,
      "hard_score": 35.7796,
      "hard_n": 71,
      "ext_overall": 67.7288,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 148,
      "class_rank": 50,
      "rank_lo": 116,
      "rank_hi": 161,
      "class_rank_lo": 35,
      "class_rank_hi": 55,
      "categories": {
        "reasoning": 65.7143,
        "math": 60.0,
        "coding": 48.6785,
        "instruction": 69.6111,
        "knowledge": 60.0,
        "long_context": 72.2222,
        "multilingual": 45.0,
        "tool_use": 85.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4868,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1667,
          "delta_mean": 1.0667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6961,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7333,
          "delta_mean": 1.5,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.45,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0857,
          "delta_mean": 0.4857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.7917,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.0246,
        "decode_tps_c4": 28.2607,
        "ttft_s": 0.4145,
        "prefill_tps_4k": 252.2024,
        "prefill_tps_12k": 217.1607,
        "load_seconds": 135.297,
        "size_gb": 22.231,
        "vram_gb": 22.231,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-3-5-mini-instruct",
      "display_name": "Phi-3.5-mini-instruct",
      "family": "phi",
      "org": "microsoft",
      "params_b": 3.82,
      "arch": "dense",
      "hf_downloads": 365509,
      "hf_rank": 420,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi3.5:3.8b-mini-instruct-q4_K_M",
      "size_bytes": 2393232384,
      "model_version_id": 21,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 67.0529,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 5.409,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9974,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.074,
      "agreement": "high",
      "disputed": 21,
      "tokens_per_s": 7.6548,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 62.03,
      "ci_high": 72.31,
      "hard_score": 20.9302,
      "hard_n": 38,
      "ext_overall": 66.9295,
      "n_truncated": 10,
      "n_extended": 10,
      "n_gated": 0,
      "beats": [
        17,
        27,
        70,
        73,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        91,
        92,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        221,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 149,
      "class_rank": 16,
      "rank_lo": 110,
      "rank_hi": 174,
      "class_rank_lo": 9,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 60.0,
        "math": 48.0,
        "coding": 38.9206,
        "instruction": 60.1667,
        "knowledge": 90.0,
        "long_context": 72.2222,
        "multilingual": 46.6667,
        "tool_use": 87.5,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3892,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6017,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4667,
          "delta_mean": 1.9,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6286,
          "delta_mean": 0.8286,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.7917,
          "delta_mean": 1.5417,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.5,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.5846,
        "decode_tps_c4": 41.5979,
        "ttft_s": 0.2837,
        "prefill_tps_4k": 276.1487,
        "prefill_tps_12k": 156.0935,
        "load_seconds": 219.1701,
        "size_gb": 28.4257,
        "vram_gb": 28.4257,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-1b-h",
      "display_name": "Granite4 1B H",
      "family": "granite",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q8_0",
      "ollama_ref": "granite4:1b-h",
      "size_bytes": 1600000000,
      "model_version_id": 123,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 66.7905,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.8488,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.0786,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1812,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 5.313,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 61.41,
      "ci_high": 72.11,
      "hard_score": 37.66,
      "hard_n": 71,
      "ext_overall": 66.9757,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        101,
        103,
        108,
        109,
        110,
        112,
        113,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        140,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        162,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 150,
      "class_rank": 11,
      "rank_lo": 119,
      "rank_hi": 162,
      "class_rank_lo": 6,
      "class_rank_hi": 12,
      "categories": {
        "reasoning": 48.5714,
        "math": 52.0,
        "coding": 59.8762,
        "instruction": 85.1111,
        "knowledge": 76.6667,
        "long_context": 55.5556,
        "multilingual": 43.3333,
        "tool_use": 80.0,
        "safety": 100.0,
        "sycophancy": 75.0,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5988,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8667,
          "delta_mean": 1.0667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8511,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.3,
          "delta_mean": 1.5,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1429,
          "delta_mean": 0.7143,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4583,
          "delta_mean": 1.375,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 24.9159,
        "decode_tps_c4": 50.142,
        "ttft_s": 0.3467,
        "prefill_tps_4k": 615.6944,
        "prefill_tps_12k": 631.2839,
        "load_seconds": 15.75,
        "size_gb": 3.3565,
        "vram_gb": 3.3565,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-math-7b-instruct",
      "display_name": "Qwen2 Math 7B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2-math-tmplfix:7b-instruct-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 194,
      "run_id": "r20261028-bv1-ef0895",
      "battery_version": "v1",
      "tested_at": "2026-10-28T10:42:23.529Z",
      "overall": 65.956,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.2262,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.5574,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0431,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 4.4507,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": null,
      "ci_low": 60.52,
      "ci_high": 71.33,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 66.1411,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 6,
      "beats": [
        17,
        73,
        75,
        76,
        77,
        84,
        85,
        86,
        87,
        88,
        91,
        92,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 151,
      "class_rank": 51,
      "rank_lo": 108,
      "rank_hi": 178,
      "class_rank_lo": 33,
      "class_rank_hi": 62,
      "categories": {
        "reasoning": 71.4286,
        "math": 76.0,
        "coding": 39.1195,
        "instruction": 68.7222,
        "knowledge": 66.6667,
        "long_context": 33.3333,
        "multilingual": 77.5,
        "tool_use": 90.0,
        "safety": 70.8333,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3912,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4667,
          "delta_mean": 0.6667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6872,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0667,
          "delta_mean": 2.0667,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6286,
          "delta_mean": 0.8571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.1515,
        "decode_tps_c4": 19.3901,
        "ttft_s": 0.3626,
        "prefill_tps_4k": 306.9087,
        "prefill_tps_12k": 319.7081,
        "load_seconds": 18.2983,
        "size_gb": 4.6831,
        "vram_gb": 4.6831,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "exaone-deep-7-8b",
      "display_name": "Exaone Deep 7.8B",
      "family": "exaone",
      "org": null,
      "params_b": 7.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "exaone-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "exaone-deep:7.8b-q4_K_M",
      "size_bytes": 4800000000,
      "model_version_id": 93,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 65.7065,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 4.3024,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8955,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0155,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 4.0555,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 59.95,
      "ci_high": 71.45,
      "hard_score": 39.8498,
      "hard_n": 71,
      "ext_overall": 68.8444,
      "n_truncated": 42,
      "n_extended": 41,
      "n_gated": 2,
      "beats": [
        17,
        70,
        73,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        118,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 152,
      "class_rank": 52,
      "rank_lo": 120,
      "rank_hi": 170,
      "class_rank_lo": 38,
      "class_rank_hi": 59,
      "categories": {
        "reasoning": 85.7143,
        "math": 88.0,
        "coding": 55.4222,
        "instruction": 16.9444,
        "knowledge": 86.6667,
        "long_context": 61.1111,
        "multilingual": 50.8333,
        "tool_use": 80.0,
        "safety": 66.6667,
        "sycophancy": 91.6667,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5542,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8333,
          "delta_mean": 0.3333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1694,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9667,
          "delta_mean": 0.9333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5143,
          "delta_mean": 0.8,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.913,
          "delta_mean": 1.2609,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.25,
          "delta_mean": 1.75,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.1798,
        "decode_tps_c4": 28.8573,
        "ttft_s": 0.3663,
        "prefill_tps_4k": 275.5567,
        "prefill_tps_12k": 229.5328,
        "load_seconds": 131.422,
        "size_gb": 22.2641,
        "vram_gb": 22.2641,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "olmo2-7b-1124-instruct",
      "display_name": "Olmo2 7B 1124 Instruct",
      "family": "olmo",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "olmo2:7b-1124-instruct-q4_K_M",
      "size_bytes": 4500000000,
      "model_version_id": 144,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 65.361,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 5.9548,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.5002,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1126,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 3.0628,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 60.57,
      "ci_high": 70.13,
      "hard_score": 23.5455,
      "hard_n": 71,
      "ext_overall": 65.361,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        74,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        118,
        120,
        121,
        124,
        128,
        133,
        134,
        142,
        145,
        148,
        152,
        153,
        154,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        177,
        178,
        179,
        180,
        181,
        184,
        186,
        187,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 153,
      "class_rank": 53,
      "rank_lo": 125,
      "rank_hi": 167,
      "class_rank_lo": 40,
      "class_rank_hi": 57,
      "categories": {
        "reasoning": 51.4286,
        "math": 52.0,
        "coding": 29.3204,
        "instruction": 84.3889,
        "knowledge": 93.3333,
        "long_context": 27.7778,
        "multilingual": 60.0,
        "tool_use": 90.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2932,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8,
          "delta_mean": 1.3667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8439,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4333,
          "delta_mean": 1.1333,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1429,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 8.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 8.25,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.7869,
        "decode_tps_c4": 16.8863,
        "ttft_s": 0.3674,
        "prefill_tps_4k": 260.5109,
        "prefill_tps_12k": 289.2235,
        "load_seconds": 112.265,
        "size_gb": 21.6279,
        "vram_gb": 21.6279,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-3b-instruct",
      "display_name": "Falcon3 3B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3-tmplfix:3b-instruct-q4_K_M",
      "size_bytes": 2000000000,
      "model_version_id": 191,
      "run_id": "r20261028-bv1-7ce712",
      "battery_version": "v1",
      "tested_at": "2026-10-28T02:36:44.578Z",
      "overall": 64.8494,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 5.4988,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.1369,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0076,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 8.2882,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 59.42,
      "ci_high": 70.62,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 65.1668,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 2,
      "beats": [
        17,
        73,
        75,
        76,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 154,
      "class_rank": 17,
      "rank_lo": 113,
      "rank_hi": 181,
      "class_rank_lo": 10,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 51.4286,
        "math": 52.0,
        "coding": 40.8823,
        "instruction": 76.8333,
        "knowledge": 73.3333,
        "long_context": 66.6667,
        "multilingual": 50.8333,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 66.6667,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4088,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5333,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7683,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 1.4667,
          "disputed": 9,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1429,
          "delta_mean": 0.7714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6667,
          "delta_mean": 1.625,
          "disputed": 5,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 0.375,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 27.9215,
        "decode_tps_c4": 60.3082,
        "ttft_s": 0.1543,
        "prefill_tps_4k": 673.5075,
        "prefill_tps_12k": 517.9889,
        "load_seconds": 82.8395,
        "size_gb": 2.0057,
        "vram_gb": 2.0057,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-2-2b-instruct",
      "display_name": "Granite3.2 2B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.2:2b-instruct-q4_K_M",
      "size_bytes": 1500000000,
      "model_version_id": 114,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 63.844,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 5.5483,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.0821,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9805,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 29.6776,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 58.73,
      "ci_high": 69.03,
      "hard_score": 23.9187,
      "hard_n": 71,
      "ext_overall": 63.844,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 1,
      "beats": [
        17,
        70,
        73,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        90,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        118,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        172,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        184,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 155,
      "class_rank": 12,
      "rank_lo": 132,
      "rank_hi": 172,
      "class_rank_lo": 6,
      "class_rank_hi": 15,
      "categories": {
        "reasoning": 42.8571,
        "math": 52.0,
        "coding": 48.4055,
        "instruction": 75.5,
        "knowledge": 80.0,
        "long_context": 50.0,
        "multilingual": 43.3333,
        "tool_use": 95.0,
        "safety": 87.5,
        "sycophancy": 75.0,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4841,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2667,
          "delta_mean": 0.9333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.755,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4,
          "delta_mean": 1.5,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2857,
          "delta_mean": 0.8857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5833,
          "delta_mean": 0.9583,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 30.0715,
        "decode_tps_c4": 32.1347,
        "ttft_s": 0.1835,
        "prefill_tps_4k": 635.3461,
        "prefill_tps_12k": 537.8871,
        "load_seconds": 71.4919,
        "size_gb": 12.6166,
        "vram_gb": 12.6166,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "swiss-ai-apertus-8b-instruct-2509",
      "display_name": "Apertus-8B-Instruct-2509",
      "family": "other",
      "org": "swiss-ai",
      "params_b": 8.05,
      "arch": "dense",
      "hf_downloads": 453286,
      "hf_rank": 337,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "apertus-tmplfix:8b-instruct-2509-q4_K_M",
      "size_bytes": null,
      "model_version_id": 195,
      "run_id": "r20261029-bv1-9ddd5f",
      "battery_version": "v1",
      "tested_at": "2026-10-29T18:10:15.755Z",
      "overall": 63.6104,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 5.5931,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9967,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0626,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 3.492,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "Vulkan (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "Vulkan",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 58.16,
      "ci_high": 68.8,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 63.6104,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        17,
        73,
        75,
        76,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 156,
      "class_rank": 54,
      "rank_lo": 125,
      "rank_hi": 182,
      "class_rank_lo": 39,
      "class_rank_hi": 63,
      "categories": {
        "reasoning": 42.8571,
        "math": 32.0,
        "coding": 33.0253,
        "instruction": 76.2778,
        "knowledge": 83.3333,
        "long_context": 50.0,
        "multilingual": 75.0,
        "tool_use": 80.0,
        "safety": 100.0,
        "sycophancy": 83.3333,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3303,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8333,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7628,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6333,
          "delta_mean": 1.3667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.32,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6,
          "delta_mean": 0.5714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.6667,
          "delta_mean": 1.25,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.25,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.5348,
        "decode_tps_c4": 32.9398,
        "ttft_s": 0.3845,
        "prefill_tps_4k": 187.8229,
        "prefill_tps_12k": 115.3327,
        "load_seconds": 10.6447,
        "size_gb": 22.3054,
        "vram_gb": 22.3054,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-3-mini-4k-instruct",
      "display_name": "Phi-3-mini-4k-instruct",
      "family": "phi",
      "org": "microsoft",
      "params_b": 3.82,
      "arch": "dense",
      "hf_downloads": 345411,
      "hf_rank": 441,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi3:3.8b-mini-4k-instruct-q4_K_M",
      "size_bytes": 2393231808,
      "model_version_id": 22,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 63.4316,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 6.0276,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.5086,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.98,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 8.4409,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 58.33,
      "ci_high": 68.09,
      "hard_score": 20.8985,
      "hard_n": 38,
      "ext_overall": 63.6167,
      "n_truncated": 13,
      "n_extended": 13,
      "n_gated": 1,
      "beats": [
        17,
        73,
        75,
        76,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 157,
      "class_rank": 18,
      "rank_lo": 129,
      "rank_hi": 182,
      "class_rank_lo": 13,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 77.1429,
        "math": 56.0,
        "coding": 35.0746,
        "instruction": 70.7222,
        "knowledge": 96.6667,
        "long_context": 11.1111,
        "multilingual": 52.5,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 66.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3507,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0333,
          "delta_mean": 0.4333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7072,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1,
          "delta_mean": 1.5,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.56,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.525,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7429,
          "delta_mean": 0.8,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0417,
          "delta_mean": 1.0417,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.6273,
        "decode_tps_c4": 44.5007,
        "ttft_s": 0.2745,
        "prefill_tps_4k": 329.0263,
        "prefill_tps_12k": 371.1952,
        "load_seconds": 57.7854,
        "size_gb": 8.9138,
        "vram_gb": 8.9138,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "phi3-3-8b-mini-128k-instruct",
      "display_name": "Phi3 3.8B Mini 128K Instruct",
      "family": "phi",
      "org": null,
      "params_b": 3.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi3:3.8b-mini-128k-instruct-q4_K_M",
      "size_bytes": 2400000000,
      "model_version_id": 150,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 62.9269,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.3545,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.8238,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0155,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 15.8177,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 58.02,
      "ci_high": 67.65,
      "hard_score": 21.8783,
      "hard_n": 71,
      "ext_overall": 63.1121,
      "n_truncated": 10,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [
        17,
        70,
        73,
        75,
        76,
        77,
        82,
        84,
        85,
        86,
        87,
        88,
        91,
        92,
        100,
        103,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        188,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 158,
      "class_rank": 19,
      "rank_lo": 137,
      "rank_hi": 176,
      "class_rank_lo": 14,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 54.2857,
        "math": 40.0,
        "coding": 38.5564,
        "instruction": 60.1667,
        "knowledge": 96.6667,
        "long_context": 50.0,
        "multilingual": 31.6667,
        "tool_use": 95.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3856,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5667,
          "delta_mean": 0.6333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6017,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6,
          "delta_mean": 1.7667,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2857,
          "delta_mean": 0.8857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.1667,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 63.9178,
        "decode_tps_c4": 96.0871,
        "ttft_s": 0.0766,
        "prefill_tps_4k": 1122.9805,
        "prefill_tps_12k": 687.4898,
        "load_seconds": 243.844,
        "size_gb": 54.0613,
        "vram_gb": 54.0613,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codellama-34b-instruct",
      "display_name": "Codellama 34B Instruct",
      "family": "llama",
      "org": null,
      "params_b": 34.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "codellama:34b-instruct-q4_K_M",
      "size_bytes": 20220000000,
      "model_version_id": 199,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 61.217,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 5.2102,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.6426,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2405,
      "agreement": "high",
      "disputed": 21,
      "tokens_per_s": 0.8179,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 55.98,
      "ci_high": 66.32,
      "hard_score": 21.8613,
      "hard_n": 35,
      "ext_overall": 61.217,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 159,
      "class_rank": 47,
      "rank_lo": 136,
      "rank_hi": 187,
      "class_rank_lo": 47,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 42.8571,
        "math": 16.0,
        "coding": 38.7069,
        "instruction": 75.0556,
        "knowledge": 73.3333,
        "long_context": 66.6667,
        "multilingual": 50.0,
        "tool_use": 92.5,
        "safety": 95.8333,
        "sycophancy": 41.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3871,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4667,
          "delta_mean": 0.9667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7506,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4333,
          "delta_mean": 1.5667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7714,
          "delta_mean": 1.0857,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9167,
          "delta_mean": 1.2083,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.1316,
        "decode_tps_c4": 7.3637,
        "ttft_s": 1.6169,
        "prefill_tps_4k": 58.0542,
        "prefill_tps_12k": 50.7601,
        "load_seconds": 190.1519,
        "size_gb": 46.7182,
        "vram_gb": 46.7182,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "smallthinker-3b-preview",
      "display_name": "Smallthinker 3B Preview",
      "family": "other",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2.5 3B",
      "license": "qwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "smallthinker:3b-preview-q4_K_M",
      "size_bytes": 2100000000,
      "model_version_id": 173,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 61.1071,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 3.9498,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.1893,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2095,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 14.4235,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 55.77,
      "ci_high": 66.63,
      "hard_score": 25.5114,
      "hard_n": 71,
      "ext_overall": 65.6961,
      "n_truncated": 48,
      "n_extended": 48,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 160,
      "class_rank": 20,
      "rank_lo": 143,
      "rank_hi": 184,
      "class_rank_lo": 15,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 57.1429,
        "math": 68.0,
        "coding": 29.8214,
        "instruction": 27.7778,
        "knowledge": 80.0,
        "long_context": 38.8889,
        "multilingual": 76.6667,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2982,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.2667,
          "delta_mean": 1.1,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3333,
          "delta_mean": 1.4,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.68,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9714,
          "delta_mean": 0.7143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.4583,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 22.113,
        "decode_tps_c4": 53.1948,
        "ttft_s": 0.2076,
        "prefill_tps_4k": 672.4052,
        "prefill_tps_12k": 587.2243,
        "load_seconds": 46.672,
        "size_gb": 7.1606,
        "vram_gb": 7.1606,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma-7b-instruct-v1-1",
      "display_name": "Gemma 7B Instruct V1.1",
      "family": "gemma",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma:7b-instruct-v1.1-q4_K_M",
      "size_bytes": 5300000000,
      "model_version_id": 101,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 60.8402,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.336,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9121,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0976,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 2.4333,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 55.54,
      "ci_high": 66.11,
      "hard_score": 28.4767,
      "hard_n": 71,
      "ext_overall": 60.8402,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        87,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202,
        236
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 161,
      "class_rank": 55,
      "rank_lo": 147,
      "rank_hi": 183,
      "class_rank_lo": 50,
      "class_rank_hi": 63,
      "categories": {
        "reasoning": 40.0,
        "math": 24.0,
        "coding": 26.4503,
        "instruction": 76.0,
        "knowledge": 83.3333,
        "long_context": 61.1111,
        "multilingual": 56.6667,
        "tool_use": 80.0,
        "safety": 100.0,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2645,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3667,
          "delta_mean": 1.1,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5333,
          "delta_mean": 1.3667,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2857,
          "delta_mean": 0.7714,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.75,
          "delta_mean": 1.5,
          "disputed": 3,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.1814,
        "decode_tps_c4": 15.1985,
        "ttft_s": 0.4962,
        "prefill_tps_4k": 197.8626,
        "prefill_tps_12k": 199.62,
        "load_seconds": 226.094,
        "size_gb": 36.1267,
        "vram_gb": 36.1267,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-1-dense-2b-instruct",
      "display_name": "Granite3.1 Dense 2B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.1-dense:2b-instruct-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 110,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 60.5775,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 5.4462,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.1183,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9126,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 7.7326,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 54.96,
      "ci_high": 66.1,
      "hard_score": 27.0265,
      "hard_n": 71,
      "ext_overall": 60.5775,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        87,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 162,
      "class_rank": 13,
      "rank_lo": 147,
      "rank_hi": 184,
      "class_rank_lo": 11,
      "class_rank_hi": 20,
      "categories": {
        "reasoning": 45.7143,
        "math": 36.0,
        "coding": 46.594,
        "instruction": 73.5556,
        "knowledge": 76.6667,
        "long_context": 50.0,
        "multilingual": 45.0,
        "tool_use": 80.0,
        "safety": 91.6667,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4659,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2333,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7356,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3,
          "delta_mean": 1.4,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.36,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.45,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.1143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6,
          "delta_mean": 0.5714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7083,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 28.7972,
        "decode_tps_c4": 56.7879,
        "ttft_s": 0.1845,
        "prefill_tps_4k": 634.8105,
        "prefill_tps_12k": 534.1775,
        "load_seconds": 70.031,
        "size_gb": 12.641,
        "vram_gb": 12.641,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-5-coder-1-5b-instruct",
      "display_name": "Qwen2.5 Coder 1.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:1.5b-instruct-q4_K_M",
      "size_bytes": 986000000,
      "model_version_id": 162,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 60.5126,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.2236,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.2867,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2874,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 12.997,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 54.93,
      "ci_high": 66.01,
      "hard_score": 28.5745,
      "hard_n": 71,
      "ext_overall": 60.6978,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 3,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        169,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 163,
      "class_rank": 14,
      "rank_lo": 147,
      "rank_hi": 185,
      "class_rank_lo": 11,
      "class_rank_hi": 20,
      "categories": {
        "reasoning": 65.7143,
        "math": 44.0,
        "coding": 52.7879,
        "instruction": 58.7778,
        "knowledge": 60.0,
        "long_context": 66.6667,
        "multilingual": 30.8333,
        "tool_use": 82.5,
        "safety": 83.3333,
        "sycophancy": 66.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5279,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4,
          "delta_mean": 0.9333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5878,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4333,
          "delta_mean": 1.9333,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.44,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6,
          "delta_mean": 1.0286,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 1.0417,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.25,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 45.7585,
        "decode_tps_c4": 77.0294,
        "ttft_s": 0.1006,
        "prefill_tps_4k": 1221.423,
        "prefill_tps_12k": 1016.0871,
        "load_seconds": 26.8341,
        "size_gb": 5.0759,
        "vram_gb": 5.0759,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistral-7b-instruct-v0-3",
      "display_name": "Mistral 7B Instruct V0.3",
      "family": "mistral",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral:7b-instruct-v0.3-q4_K_M",
      "size_bytes": 4400000000,
      "model_version_id": 140,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 59.775,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.2438,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.6112,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1536,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 3.1401,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 54.67,
      "ci_high": 64.7,
      "hard_score": 24.9023,
      "hard_n": 71,
      "ext_overall": 59.5898,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 164,
      "class_rank": 56,
      "rank_lo": 148,
      "rank_hi": 186,
      "class_rank_lo": 50,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 37.1429,
        "math": 16.0,
        "coding": 40.8875,
        "instruction": 70.3333,
        "knowledge": 93.3333,
        "long_context": 61.1111,
        "multilingual": 41.6667,
        "tool_use": 90.0,
        "safety": 87.5,
        "sycophancy": 66.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4089,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0333,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7033,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9667,
          "delta_mean": 1.4667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5143,
          "delta_mean": 0.9429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.4167,
          "delta_mean": 1.2083,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.5781,
        "decode_tps_c4": 17.5228,
        "ttft_s": 0.4403,
        "prefill_tps_4k": 271.76,
        "prefill_tps_12k": 222.8042,
        "load_seconds": 125.86,
        "size_gb": 22.0301,
        "vram_gb": 22.0301,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-1-moe-3b-instruct",
      "display_name": "Granite3.1 Moe 3B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.1-moe:3b-instruct-q4_K_M",
      "size_bytes": 2000000000,
      "model_version_id": 113,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 59.7683,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 4.9881,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.6086,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0257,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 11.5725,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 54.66,
      "ci_high": 65.02,
      "hard_score": 24.9488,
      "hard_n": 71,
      "ext_overall": 60.3979,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 165,
      "class_rank": 21,
      "rank_lo": 148,
      "rank_hi": 186,
      "class_rank_lo": 16,
      "class_rank_hi": 21,
      "categories": {
        "reasoning": 37.1429,
        "math": 32.0,
        "coding": 32.6053,
        "instruction": 69.2222,
        "knowledge": 83.3333,
        "long_context": 61.1111,
        "multilingual": 41.6667,
        "tool_use": 85.0,
        "safety": 95.8333,
        "sycophancy": 58.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3261,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6922,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9667,
          "delta_mean": 1.5,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.32,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9429,
          "delta_mean": 0.6286,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7083,
          "delta_mean": 1.2083,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.625,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 53.1127,
        "decode_tps_c4": 92.9557,
        "ttft_s": 0.1254,
        "prefill_tps_4k": 1079.3789,
        "prefill_tps_12k": 838.4479,
        "load_seconds": 61.109,
        "size_gb": 10.9422,
        "vram_gb": 10.9422,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codegemma-7b-instruct-v1-1",
      "display_name": "Codegemma 7B Instruct V1.1",
      "family": "gemma",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "codegemma:7b-instruct-v1.1-q4_K_M",
      "size_bytes": 5300000000,
      "model_version_id": 74,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 59.5044,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 5.7514,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.3755,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9317,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 3.4929,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 54.8,
      "ci_high": 64.19,
      "hard_score": 29.3339,
      "hard_n": 71,
      "ext_overall": 59.5044,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 166,
      "class_rank": 57,
      "rank_lo": 150,
      "rank_hi": 186,
      "class_rank_lo": 51,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 45.7143,
        "math": 24.0,
        "coding": 55.5474,
        "instruction": 78.0556,
        "knowledge": 70.0,
        "long_context": 5.5556,
        "multilingual": 70.0,
        "tool_use": 95.0,
        "safety": 91.6667,
        "sycophancy": 75.0,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5555,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3667,
          "delta_mean": 1.1333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7806,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.1333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8857,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.625,
          "delta_mean": 1.2917,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.2962,
        "decode_tps_c4": 52.5964,
        "ttft_s": 0.4963,
        "prefill_tps_4k": 198.9778,
        "prefill_tps_12k": 355.9046,
        "load_seconds": 256.719,
        "size_gb": 36.1267,
        "vram_gb": 36.1267,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistralai-mistral-7b-instruct-v0-2",
      "display_name": "Mistral-7B-Instruct-v0.2",
      "family": "mistral",
      "org": "mistralai",
      "params_b": 7.24,
      "arch": "dense",
      "hf_downloads": 1660754,
      "hf_rank": 88,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "mistral:7b-instruct-v0.2-q4_K_M",
      "size_bytes": 4369375808,
      "model_version_id": 27,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 59.2136,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.0788,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7806,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9466,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 3.5885,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 54.33,
      "ci_high": 64.22,
      "hard_score": 18.7305,
      "hard_n": 38,
      "ext_overall": 59.3988,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 167,
      "class_rank": 58,
      "rank_lo": 143,
      "rank_hi": 191,
      "class_rank_lo": 47,
      "class_rank_hi": 65,
      "categories": {
        "reasoning": 40.0,
        "math": 12.0,
        "coding": 25.3112,
        "instruction": 75.0556,
        "knowledge": 90.0,
        "long_context": 72.2222,
        "multilingual": 42.5,
        "tool_use": 80.0,
        "safety": 95.8333,
        "sycophancy": 91.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2531,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3,
          "delta_mean": 0.7667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7506,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3667,
          "delta_mean": 1.5333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.12,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.425,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0571,
          "delta_mean": 0.5143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.3043,
          "delta_mean": 1.0435,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.2045,
        "decode_tps_c4": 29.0183,
        "ttft_s": 0.4304,
        "prefill_tps_4k": 274.1602,
        "prefill_tps_12k": 224.9593,
        "load_seconds": 131.8757,
        "size_gb": 21.994,
        "vram_gb": 21.994,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "yi-coder-9b-chat",
      "display_name": "YI Coder 9B Chat",
      "family": "yi",
      "org": null,
      "params_b": 9.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "yi-coder:9b-chat-q4_K_M",
      "size_bytes": 5300000000,
      "model_version_id": 187,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 59.0417,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 4.6307,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.4429,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9648,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.2091,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 53.71,
      "ci_high": 64.45,
      "hard_score": 37.7912,
      "hard_n": 71,
      "ext_overall": 59.0417,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 4,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 168,
      "class_rank": 59,
      "rank_lo": 150,
      "rank_hi": 186,
      "class_rank_lo": 51,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 31.4286,
        "math": 16.0,
        "coding": 81.2523,
        "instruction": 65.3333,
        "knowledge": 83.3333,
        "long_context": 44.4444,
        "multilingual": 54.1667,
        "tool_use": 95.0,
        "safety": 60.4167,
        "sycophancy": 75.0,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8125,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.6333,
          "delta_mean": 1.1667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6533,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7,
          "delta_mean": 1.2667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5714,
          "delta_mean": 0.4286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6042,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4348,
          "delta_mean": 1.087,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.4522,
        "decode_tps_c4": 24.6006,
        "ttft_s": 0.5387,
        "prefill_tps_4k": 218.812,
        "prefill_tps_12k": 188.8952,
        "load_seconds": 98.703,
        "size_gb": 18.5368,
        "vram_gb": 18.5368,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "solar-10-7b-instruct-v1",
      "display_name": "Solar 10.7B Instruct V1",
      "family": "solar",
      "org": null,
      "params_b": 10.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "solar:10.7b-instruct-v1-q4_K_M",
      "size_bytes": 6500000000,
      "model_version_id": 177,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 58.6833,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 5.3593,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7537,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.055,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 2.894,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 53.66,
      "ci_high": 63.79,
      "hard_score": 26.4035,
      "hard_n": 71,
      "ext_overall": 58.6833,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        84,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 169,
      "class_rank": 20,
      "rank_lo": 151,
      "rank_hi": 186,
      "class_rank_lo": 20,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 45.7143,
        "math": 16.0,
        "coding": 34.3241,
        "instruction": 69.8889,
        "knowledge": 86.6667,
        "long_context": 38.8889,
        "multilingual": 58.3333,
        "tool_use": 95.0,
        "safety": 83.3333,
        "sycophancy": 83.3333,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3432,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7667,
          "delta_mean": 1.0333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6989,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6667,
          "delta_mean": 1.5667,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5143,
          "delta_mean": 0.8,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6957,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.6758,
        "decode_tps_c4": 13.5854,
        "ttft_s": 0.609,
        "prefill_tps_4k": 182.8502,
        "prefill_tps_12k": 190.5144,
        "load_seconds": 47.25,
        "size_gb": 13.0813,
        "vram_gb": 13.0813,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-dense-2b-instruct",
      "display_name": "Granite3 Dense 2B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3-dense:2b-instruct-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 118,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 58.3099,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 5.3764,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7379,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1314,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 8.9416,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 53.08,
      "ci_high": 63.63,
      "hard_score": 20.1859,
      "hard_n": 71,
      "ext_overall": 58.3099,
      "n_truncated": 3,
      "n_extended": 3,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 170,
      "class_rank": 15,
      "rank_lo": 152,
      "rank_hi": 187,
      "class_rank_lo": 13,
      "class_rank_hi": 22,
      "categories": {
        "reasoning": 48.5714,
        "math": 24.0,
        "coding": 42.1623,
        "instruction": 63.3889,
        "knowledge": 86.6667,
        "long_context": 33.3333,
        "multilingual": 48.3333,
        "tool_use": 82.5,
        "safety": 95.8333,
        "sycophancy": 75.0,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4216,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5667,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6339,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2,
          "delta_mean": 1.6333,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7143,
          "delta_mean": 0.8571,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.8333,
          "delta_mean": 1.1667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.375,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 29.6697,
        "decode_tps_c4": 43.4875,
        "ttft_s": 0.1839,
        "prefill_tps_4k": 625.615,
        "prefill_tps_12k": 652.2115,
        "load_seconds": 20.109,
        "size_gb": 4.3705,
        "vram_gb": 4.3705,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-v2-16b-lite-chat",
      "display_name": "Deepseek V2 16B Lite Chat",
      "family": "deepseek",
      "org": null,
      "params_b": 16.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": 2.4,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-v2:16b-lite-chat-q4_K_M",
      "size_bytes": 10000000000,
      "model_version_id": 90,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 58.1802,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 5.4212,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7943,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1493,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 11.6271,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 53.15,
      "ci_high": 63.44,
      "hard_score": 20.0533,
      "hard_n": 71,
      "ext_overall": 58.7357,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 171,
      "class_rank": 21,
      "rank_lo": 152,
      "rank_hi": 188,
      "class_rank_lo": 20,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 34.2857,
        "math": 40.0,
        "coding": 41.6136,
        "instruction": 70.2222,
        "knowledge": 70.0,
        "long_context": 33.3333,
        "multilingual": 50.8333,
        "tool_use": 87.5,
        "safety": 95.8333,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4161,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4,
          "delta_mean": 1.1333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7022,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.9333,
          "delta_mean": 1.6667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9714,
          "delta_mean": 0.5714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.7917,
          "delta_mean": 1.25,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.1741,
        "decode_tps_c4": 5.9416,
        "ttft_s": 0.4913,
        "prefill_tps_4k": 244.1408,
        "prefill_tps_12k": 108.9644,
        "load_seconds": 243.39,
        "size_gb": 46.8174,
        "vram_gb": 46.8174,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "yi-6b-chat-v1-5",
      "display_name": "YI 6B Chat V1.5",
      "family": "yi",
      "org": null,
      "params_b": 6.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "yi:6b-chat-v1.5-q4_K_M",
      "size_bytes": 3700000000,
      "model_version_id": 184,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 58.0319,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 4.734,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.3512,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2005,
      "agreement": "high",
      "disputed": 19,
      "tokens_per_s": 16.3458,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 52.84,
      "ci_high": 63.52,
      "hard_score": 20.993,
      "hard_n": 71,
      "ext_overall": 57.8837,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 1,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 172,
      "class_rank": 60,
      "rank_lo": 152,
      "rank_hi": 187,
      "class_rank_lo": 52,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 45.7143,
        "math": 48.0,
        "coding": 19.7946,
        "instruction": 70.4444,
        "knowledge": 73.3333,
        "long_context": 33.3333,
        "multilingual": 51.6667,
        "tool_use": 92.5,
        "safety": 87.5,
        "sycophancy": 75.0,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1979,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.8,
          "delta_mean": 0.5667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7044,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.8333,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3143,
          "delta_mean": 0.6857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 1.4167,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.925,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.5,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 43.3222,
        "decode_tps_c4": 120.2442,
        "ttft_s": 0.1149,
        "prefill_tps_4k": 1256.1834,
        "prefill_tps_12k": 1353.4998,
        "load_seconds": 16.859,
        "size_gb": 5.8833,
        "vram_gb": 5.8833,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "llama3-2-1b-instruct",
      "display_name": "Llama3.2 1B Instruct",
      "family": "llama",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama3.2",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama3.2:1b-instruct-q4_K_M",
      "size_bytes": 808000000,
      "model_version_id": 134,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 58.0189,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 13,
      "judge_a_mean": 5.2214,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.5471,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1338,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 17.1378,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 52.29,
      "ci_high": 63.72,
      "hard_score": 26.582,
      "hard_n": 71,
      "ext_overall": 58.0189,
      "n_truncated": 4,
      "n_extended": 3,
      "n_gated": 3,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        142,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 173,
      "class_rank": 16,
      "rank_lo": 150,
      "rank_hi": 187,
      "class_rank_lo": 12,
      "class_rank_hi": 22,
      "categories": {
        "reasoning": 37.1429,
        "math": 40.0,
        "coding": 41.6937,
        "instruction": 76.9444,
        "knowledge": 76.6667,
        "long_context": 38.8889,
        "multilingual": 55.0,
        "tool_use": 72.5,
        "safety": 83.3333,
        "sycophancy": 50.0,
        "gotchas": 13.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4169,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6333,
          "delta_mean": 1.0333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.1333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7694,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4667,
          "delta_mean": 1.5333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.55,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8857,
          "delta_mean": 0.6857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.75,
          "delta_mean": 1.0417,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.725,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 58.5414,
        "decode_tps_c4": 114.5358,
        "ttft_s": 0.0685,
        "prefill_tps_4k": 1602.3985,
        "prefill_tps_12k": 1381.4463,
        "load_seconds": 34.6337,
        "size_gb": 5.4304,
        "vram_gb": 5.4304,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "sailor2-8b-chat",
      "display_name": "Sailor2 8B Chat",
      "family": "other",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2.5",
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "sailor2:8b-chat-q4_K_M",
      "size_bytes": 5200000000,
      "model_version_id": 172,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 57.2899,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 4.5166,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.2505,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0789,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 3.5114,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 51.68,
      "ci_high": 62.82,
      "hard_score": 21.1161,
      "hard_n": 71,
      "ext_overall": 57.2899,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 174,
      "class_rank": 61,
      "rank_lo": 152,
      "rank_hi": 188,
      "class_rank_lo": 52,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 54.2857,
        "math": 52.0,
        "coding": 19.4621,
        "instruction": 52.2222,
        "knowledge": 80.0,
        "long_context": 55.5556,
        "multilingual": 81.6667,
        "tool_use": 35.0,
        "safety": 85.4167,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1946,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6667,
          "delta_mean": 0.3667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0333,
          "delta_mean": 1.6667,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.5429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5143,
          "delta_mean": 0.7143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8542,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4783,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.913,
          "delta_mean": 1.5217,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 5,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.125,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.5903,
        "decode_tps_c4": 26.2881,
        "ttft_s": 0.4249,
        "prefill_tps_4k": 223.2833,
        "prefill_tps_12k": 204.3435,
        "load_seconds": 86.312,
        "size_gb": 14.1708,
        "vram_gb": 14.1708,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma3-1b-it",
      "display_name": "Gemma3 1B IT",
      "family": "gemma",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma3:1b-it-q4_K_M",
      "size_bytes": 806000000,
      "model_version_id": 103,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 56.8376,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 5.4462,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.9479,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1483,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 20.8072,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 51.76,
      "ci_high": 62.15,
      "hard_score": 29.3396,
      "hard_n": 71,
      "ext_overall": 56.5201,
      "n_truncated": 22,
      "n_extended": 22,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        158,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 175,
      "class_rank": 17,
      "rank_lo": 154,
      "rank_hi": 188,
      "class_rank_lo": 13,
      "class_rank_hi": 22,
      "categories": {
        "reasoning": 60.0,
        "math": 64.0,
        "coding": 42.538,
        "instruction": 71.9444,
        "knowledge": 63.3333,
        "long_context": 22.2222,
        "multilingual": 52.5,
        "tool_use": 35.0,
        "safety": 100.0,
        "sycophancy": 58.3333,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4254,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5333,
          "delta_mean": 1.0667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7194,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.8333,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.64,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.525,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3143,
          "delta_mean": 0.8,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 1.0417,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.875,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 56.9996,
        "decode_tps_c4": 80.0814,
        "ttft_s": 0.1024,
        "prefill_tps_4k": 1928.8723,
        "prefill_tps_12k": 1987.208,
        "load_seconds": 11.268,
        "size_gb": 1.5463,
        "vram_gb": 1.5463,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mixtral-8x7b-instruct-v0-1",
      "display_name": "Mixtral 8x7B Instruct V0.1",
      "family": "mistral",
      "org": null,
      "params_b": 56.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "mixtral:8x7b-instruct-v0.1-q4_K_M",
      "size_bytes": 28448000000,
      "model_version_id": 221,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 56.7665,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 4.7512,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.5879,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9826,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 7.4243,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "gt35",
      "size_class_label": "> 35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 51.32,
      "ci_high": 61.98,
      "hard_score": 17.4683,
      "hard_n": 35,
      "ext_overall": 56.7665,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        17,
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        124,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 176,
      "class_rank": 8,
      "rank_lo": 150,
      "rank_hi": 194,
      "class_rank_lo": 8,
      "class_rank_hi": 8,
      "categories": {
        "reasoning": 42.8571,
        "math": 24.0,
        "coding": 20.0968,
        "instruction": 62.5556,
        "knowledge": 83.3333,
        "long_context": 72.2222,
        "multilingual": 41.6667,
        "tool_use": 85.0,
        "safety": 79.1667,
        "sycophancy": 83.3333,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.201,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.0,
          "delta_mean": 0.7667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6256,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3,
          "delta_mean": 1.4333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.7222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5143,
          "delta_mean": 0.8,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0,
          "delta_mean": 0.913,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.125,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 7.5847,
        "decode_tps_c4": 8.494,
        "ttft_s": 1.3171,
        "prefill_tps_4k": 136.8653,
        "prefill_tps_12k": 127.0171,
        "load_seconds": 139.9958,
        "size_gb": 46.4599,
        "vram_gb": 46.4599,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "aya-8b-23",
      "display_name": "Aya 8B 23",
      "family": "cohere",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "cc-by-nc-4.0",
      "quantization": "F16",
      "ollama_ref": "aya:8b-23-q4_K_M",
      "size_bytes": 5100000000,
      "model_version_id": 70,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 55.9173,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.8471,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.3917,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9469,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 2.8419,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 51.13,
      "ci_high": 60.71,
      "hard_score": 24.9213,
      "hard_n": 71,
      "ext_overall": 56.0459,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        73,
        77,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        128,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 177,
      "class_rank": 62,
      "rank_lo": 155,
      "rank_hi": 189,
      "class_rank_lo": 53,
      "class_rank_hi": 64,
      "categories": {
        "reasoning": 34.2857,
        "math": 0.0,
        "coding": 30.6918,
        "instruction": 74.6667,
        "knowledge": 76.6667,
        "long_context": 61.1111,
        "multilingual": 39.1667,
        "tool_use": 95.0,
        "safety": 91.6667,
        "sycophancy": 66.6667,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3069,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5333,
          "delta_mean": 1.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7467,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.4333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8,
          "delta_mean": 0.7429,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.375,
          "delta_mean": 1.0833,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.375,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.9136,
        "decode_tps_c4": 16.4003,
        "ttft_s": 0.4043,
        "prefill_tps_4k": 243.0161,
        "prefill_tps_12k": 245.7259,
        "load_seconds": 84.516,
        "size_gb": 14.1562,
        "vram_gb": 14.1562,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepcoder-1-5b-preview",
      "display_name": "Deepcoder 1.5B Preview",
      "family": "other",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "DeepSeek-R1-Distill-Qwen 1.5B",
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepcoder:1.5b-preview-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 82,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 55.8294,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 3,
      "judge_a_mean": 4.3,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.597,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9496,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 13.0495,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 50.28,
      "ci_high": 61.03,
      "hard_score": 35.2198,
      "hard_n": 71,
      "ext_overall": 62.8982,
      "n_truncated": 47,
      "n_extended": 47,
      "n_gated": 11,
      "beats": [
        17,
        73,
        85,
        86,
        88,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 178,
      "class_rank": 18,
      "rank_lo": 155,
      "rank_hi": 192,
      "class_rank_lo": 13,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 80.0,
        "math": 88.0,
        "coding": 34.2424,
        "instruction": 68.5556,
        "knowledge": 56.6667,
        "long_context": 16.6667,
        "multilingual": 65.8333,
        "tool_use": 67.5,
        "safety": 25.0,
        "sycophancy": 66.6667,
        "gotchas": 80.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3424,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3333,
          "delta_mean": 0.1667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6856,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8,
          "delta_mean": 1.9,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.88,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.9714,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.1304,
          "delta_mean": 0.9565,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 2,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.675,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.75,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 44.8693,
        "decode_tps_c4": 66.8844,
        "ttft_s": 0.1035,
        "prefill_tps_4k": 1294.875,
        "prefill_tps_12k": 1006.5976,
        "load_seconds": 26.7731,
        "size_gb": 5.0759,
        "vram_gb": 5.0759,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-14b-chat-v1-5",
      "display_name": "Qwen 14B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 14.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen:14b-chat-v1.5-q4_K_M",
      "size_bytes": 9200000000,
      "model_version_id": 154,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 54.7805,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 5.0067,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.5983,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8445,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 2.8336,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 49.81,
      "ci_high": 59.43,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 54.7805,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [
        17,
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        190,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 179,
      "class_rank": 22,
      "rank_lo": 154,
      "rank_hi": 195,
      "class_rank_lo": 20,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 48.5714,
        "math": 40.0,
        "coding": 9.7306,
        "instruction": 76.9444,
        "knowledge": 93.3333,
        "long_context": 11.1111,
        "multilingual": 52.5,
        "tool_use": 65.0,
        "safety": 95.8333,
        "sycophancy": 66.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0973,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9,
          "delta_mean": 0.1667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7694,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8333,
          "delta_mean": 1.1333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.525,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.0833,
          "delta_mean": 1.3333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.375,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.2729,
        "decode_tps_c4": 13.9018,
        "ttft_s": 0.7251,
        "prefill_tps_4k": 136.4316,
        "prefill_tps_12k": 104.5296,
        "load_seconds": 462.547,
        "size_gb": 62.5771,
        "vram_gb": 62.5771,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "exaone-deep-2-4b",
      "display_name": "Exaone Deep 2.4B",
      "family": "exaone",
      "org": null,
      "params_b": 2.4,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "exaone-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "exaone-deep:2.4b-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 92,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 54.6755,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 1,
      "judge_a_mean": 3.5957,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.1948,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9273,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 9.2073,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 48.91,
      "ci_high": 60.46,
      "hard_score": 23.0047,
      "hard_n": 71,
      "ext_overall": 62.594,
      "n_truncated": 72,
      "n_extended": 71,
      "n_gated": 3,
      "beats": [
        17,
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        124,
        145,
        148,
        152,
        153,
        155,
        157,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 180,
      "class_rank": 19,
      "rank_lo": 156,
      "rank_hi": 196,
      "class_rank_lo": 13,
      "class_rank_hi": 25,
      "categories": {
        "reasoning": 82.8571,
        "math": 72.0,
        "coding": 36.6667,
        "instruction": 13.8889,
        "knowledge": 63.3333,
        "long_context": 50.0,
        "multilingual": 38.3333,
        "tool_use": 72.5,
        "safety": 62.5,
        "sycophancy": 75.0,
        "gotchas": 93.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.6667,
          "delta_mean": 0.7,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.9333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1389,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.7333,
          "delta_mean": 1.1667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.8286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.0286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.6286,
          "delta_mean": 0.5143,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6957,
          "delta_mean": 1.1304,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.725,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.25,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 31.1631,
        "decode_tps_c4": 65.082,
        "ttft_s": 0.1438,
        "prefill_tps_4k": 776.836,
        "prefill_tps_12k": 615.3835,
        "load_seconds": 76.516,
        "size_gb": 11.8868,
        "vram_gb": 11.8868,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-llm-7b-chat",
      "display_name": "Deepseek Llm 7B Chat",
      "family": "deepseek",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-llm:7b-chat-q4_K_M",
      "size_bytes": 4200000000,
      "model_version_id": 87,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 54.3309,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.9605,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.2714,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1414,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 3.1669,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 49.25,
      "ci_high": 59.81,
      "hard_score": 20.8524,
      "hard_n": 71,
      "ext_overall": 54.3309,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [
        17,
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        121,
        124,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        159,
        161,
        165,
        171,
        174,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 181,
      "class_rank": 63,
      "rank_lo": 158,
      "rank_hi": 194,
      "class_rank_lo": 55,
      "class_rank_hi": 65,
      "categories": {
        "reasoning": 37.1429,
        "math": 24.0,
        "coding": 37.3911,
        "instruction": 66.8333,
        "knowledge": 76.6667,
        "long_context": 27.7778,
        "multilingual": 45.8333,
        "tool_use": 77.5,
        "safety": 95.8333,
        "sycophancy": 25.0,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3739,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0667,
          "delta_mean": 1.2,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6683,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4,
          "delta_mean": 1.7667,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0571,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.2083,
          "delta_mean": 1.0833,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.2261,
        "decode_tps_c4": 24.8056,
        "ttft_s": 0.3661,
        "prefill_tps_4k": 278.355,
        "prefill_tps_12k": 299.0783,
        "load_seconds": 118.078,
        "size_gb": 20.3004,
        "vram_gb": 20.3004,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen3-vl-2b-thinking",
      "display_name": "Qwen3 VL 2B Thinking",
      "family": "qwen",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": true,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3-vl:2b-thinking-q4_K_M",
      "size_bytes": 1900000000,
      "model_version_id": 169,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 53.7037,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 3.7424,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.324,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.5844,
      "agreement": "high",
      "disputed": 6,
      "tokens_per_s": 25.7355,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 48.25,
      "ci_high": 59.13,
      "hard_score": 26.7371,
      "hard_n": 71,
      "ext_overall": 59.3757,
      "n_truncated": 106,
      "n_extended": 106,
      "n_gated": 1,
      "beats": [
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        124,
        145,
        148,
        152,
        153,
        155,
        157,
        159,
        161,
        165,
        171,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 182,
      "class_rank": 20,
      "rank_lo": 161,
      "rank_hi": 198,
      "class_rank_lo": 15,
      "class_rank_hi": 26,
      "categories": {
        "reasoning": 60.0,
        "math": 52.0,
        "coding": 0.0,
        "instruction": 77.1667,
        "knowledge": 73.3333,
        "long_context": 33.3333,
        "multilingual": 63.3333,
        "tool_use": 82.5,
        "safety": 41.6667,
        "sycophancy": 50.0,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.0,
          "delta_mean": 0.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7717,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5667,
          "delta_mean": 1.0667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.52,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.4286,
          "delta_mean": 0.5143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8182,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5,
          "delta_mean": 0.5909,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 4,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.125,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 27.151,
        "decode_tps_c4": 26.4367,
        "ttft_s": 0.1442,
        "prefill_tps_4k": 949.6726,
        "prefill_tps_12k": 806.4299,
        "load_seconds": 19.016,
        "size_gb": 3.5949,
        "vram_gb": 3.5949,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codellama-7b-instruct",
      "display_name": "Codellama 7B Instruct",
      "family": "llama",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama2",
      "quantization": "Q4_K_M",
      "ollama_ref": "codellama:7b-instruct-q4_K_M",
      "size_bytes": 4100000000,
      "model_version_id": 76,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 53.0242,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.7367,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.2471,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2252,
      "agreement": "high",
      "disputed": 18,
      "tokens_per_s": 5.2298,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 47.8,
      "ci_high": 57.77,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 52.839,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 3,
      "beats": [
        73,
        85,
        86,
        100,
        108,
        109,
        120,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 183,
      "class_rank": 64,
      "rank_lo": 159,
      "rank_hi": 202,
      "class_rank_lo": 55,
      "class_rank_hi": 66,
      "categories": {
        "reasoning": 37.1429,
        "math": 4.0,
        "coding": 36.0192,
        "instruction": 68.1111,
        "knowledge": 76.6667,
        "long_context": 61.1111,
        "multilingual": 21.6667,
        "tool_use": 85.0,
        "safety": 87.5,
        "sycophancy": 50.0,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3602,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8667,
          "delta_mean": 1.3,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6811,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6,
          "delta_mean": 1.6667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.04,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6857,
          "delta_mean": 0.7429,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7083,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.375,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.3667,
        "decode_tps_c4": 27.182,
        "ttft_s": 0.4032,
        "prefill_tps_4k": 252.3193,
        "prefill_tps_12k": 181.3559,
        "load_seconds": 248.406,
        "size_gb": 38.8382,
        "vram_gb": 38.8382,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codellama-13b-instruct",
      "display_name": "Codellama 13B Instruct",
      "family": "llama",
      "org": null,
      "params_b": 13.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama2",
      "quantization": "Q4_K_M",
      "ollama_ref": "codellama:13b-instruct-q4_K_M",
      "size_bytes": 7900000000,
      "model_version_id": 75,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 52.4674,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 4.5805,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.0424,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8733,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.0149,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 47.15,
      "ci_high": 57.79,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 52.4674,
      "n_truncated": 11,
      "n_extended": 11,
      "n_gated": 4,
      "beats": [
        73,
        85,
        86,
        108,
        109,
        120,
        133,
        145,
        148,
        152,
        153,
        155,
        156,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 184,
      "class_rank": 23,
      "rank_lo": 159,
      "rank_hi": 203,
      "class_rank_lo": 20,
      "class_rank_hi": 23,
      "categories": {
        "reasoning": 31.4286,
        "math": 12.0,
        "coding": 22.5,
        "instruction": 73.7778,
        "knowledge": 83.3333,
        "long_context": 50.0,
        "multilingual": 33.3333,
        "tool_use": 82.5,
        "safety": 83.3333,
        "sycophancy": 50.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.225,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.9,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7378,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3,
          "delta_mean": 1.4,
          "disputed": 9,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.12,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4286,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.8333,
          "delta_mean": 0.7917,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.75,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 6.9338,
        "decode_tps_c4": 16.1818,
        "ttft_s": 0.7033,
        "prefill_tps_4k": 121.7204,
        "prefill_tps_12k": 97.7222,
        "load_seconds": 444.437,
        "size_gb": 61.7475,
        "vram_gb": 61.7475,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepscaler-1-5b-preview",
      "display_name": "Deepscaler 1.5B Preview",
      "family": "other",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "DeepSeek-R1-Distill-Qwen 1.5B",
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepscaler:1.5b-preview-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 84,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 52.0944,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 3.8052,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.2274,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7731,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 44.0453,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": null,
      "ci_low": 46.62,
      "ci_high": 57.72,
      "hard_score": 30.3756,
      "hard_n": 71,
      "ext_overall": 55.3563,
      "n_truncated": 60,
      "n_extended": 58,
      "n_gated": 9,
      "beats": [
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        124,
        145,
        148,
        152,
        153,
        155,
        157,
        159,
        161,
        165,
        171,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 185,
      "class_rank": 21,
      "rank_lo": 168,
      "rank_hi": 198,
      "class_rank_lo": 15,
      "class_rank_hi": 26,
      "categories": {
        "reasoning": 68.5714,
        "math": 92.0,
        "coding": 26.6667,
        "instruction": 66.8889,
        "knowledge": 50.0,
        "long_context": 22.2222,
        "multilingual": 56.6667,
        "tool_use": 65.0,
        "safety": 20.8333,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.7,
          "delta_mean": 0.2,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6689,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9667,
          "delta_mean": 1.4667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.92,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.2083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.0952,
          "delta_mean": 0.6667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 4,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.375,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 45.5302,
        "decode_tps_c4": 40.0417,
        "ttft_s": 0.101,
        "prefill_tps_4k": 1287.3196,
        "prefill_tps_12k": 1006.4777,
        "load_seconds": 29.1182,
        "size_gb": 5.0759,
        "vram_gb": 5.0759,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "nemotron-mini-4b-instruct",
      "display_name": "Nemotron Mini 4B Instruct",
      "family": "nemotron",
      "org": null,
      "params_b": 4.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "nemotron-mini:4b-instruct-q4_K_M",
      "size_bytes": 2700000000,
      "model_version_id": 142,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 51.9667,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 4.5945,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.9221,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.04,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 21.6957,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 47.38,
      "ci_high": 56.88,
      "hard_score": 22.3712,
      "hard_n": 71,
      "ext_overall": 51.9667,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 1,
      "beats": [
        73,
        85,
        86,
        91,
        100,
        108,
        109,
        112,
        120,
        124,
        145,
        148,
        152,
        153,
        155,
        157,
        159,
        161,
        165,
        171,
        175,
        176,
        178,
        179,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 186,
      "class_rank": 22,
      "rank_lo": 171,
      "rank_hi": 198,
      "class_rank_lo": 22,
      "class_rank_hi": 24,
      "categories": {
        "reasoning": 34.2857,
        "math": 12.0,
        "coding": 34.5816,
        "instruction": 72.3889,
        "knowledge": 83.3333,
        "long_context": 11.1111,
        "multilingual": 47.5,
        "tool_use": 85.0,
        "safety": 87.5,
        "sycophancy": 50.0,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3458,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7333,
          "delta_mean": 0.9,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7239,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.6333,
          "delta_mean": 1.8667,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.12,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.475,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2857,
          "delta_mean": 0.6,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0833,
          "delta_mean": 1.3333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.85,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 66.8076,
        "decode_tps_c4": 167.5941,
        "ttft_s": 0.0601,
        "prefill_tps_4k": 2233.8113,
        "prefill_tps_12k": 2567.3169,
        "load_seconds": 21.109,
        "size_gb": 6.6401,
        "vram_gb": 6.6401,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "internlm2-1-8b-chat-v2-5",
      "display_name": "Internlm2 1.8B Chat V2.5",
      "family": "internlm",
      "org": null,
      "params_b": 1.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "internlm2-tmplfix:1.8b-chat-v2.5-q4_K_M",
      "size_bytes": 1200000000,
      "model_version_id": 188,
      "run_id": "r20261028-bv1-2e24cd",
      "battery_version": "v1",
      "tested_at": "2026-10-28T09:22:13.935Z",
      "overall": 51.8842,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.6962,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.7879,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2283,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 11.5984,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 46.05,
      "ci_high": 57.48,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 52.2546,
      "n_truncated": 5,
      "n_extended": 5,
      "n_gated": 4,
      "beats": [
        73,
        85,
        86,
        108,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 187,
      "class_rank": 22,
      "rank_lo": 159,
      "rank_hi": 207,
      "class_rank_lo": 13,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 31.4286,
        "math": 28.0,
        "coding": 48.4739,
        "instruction": 56.5556,
        "knowledge": 70.0,
        "long_context": 50.0,
        "multilingual": 29.1667,
        "tool_use": 70.0,
        "safety": 83.3333,
        "sycophancy": 41.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4847,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1,
          "delta_mean": 0.9667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5656,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6,
          "delta_mean": 1.7,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.28,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1143,
          "delta_mean": 0.6,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.25,
          "delta_mean": 1.375,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 45.5755,
        "decode_tps_c4": 85.5815,
        "ttft_s": 0.0987,
        "prefill_tps_4k": 1085.6514,
        "prefill_tps_12k": 889.1267,
        "load_seconds": 82.2355,
        "size_gb": 1.1724,
        "vram_gb": 1.1724,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "mistralai-magistral-small-2506",
      "display_name": "Magistral Small (2506)",
      "family": "mistral",
      "org": "mistralai",
      "params_b": 24.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "magistral:24b-small-2506-q4_K_M",
      "size_bytes": 14333907488,
      "model_version_id": 236,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": 49.7972,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 11,
      "judge_a_mean": 2.7193,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.465,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.4352,
      "agreement": "high",
      "disputed": 2,
      "tokens_per_s": 1.6507,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 44.54,
      "ci_high": 55.3,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 65.8385,
      "n_truncated": 119,
      "n_extended": 89,
      "n_gated": 0,
      "beats": [
        73,
        85,
        86,
        108,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 188,
      "class_rank": 48,
      "rank_lo": 162,
      "rank_hi": 207,
      "class_rank_lo": 48,
      "class_rank_hi": 48,
      "categories": {
        "reasoning": 34.2857,
        "math": 48.0,
        "coding": 10.0,
        "instruction": 34.7778,
        "knowledge": 56.6667,
        "long_context": 44.4444,
        "multilingual": 70.0,
        "tool_use": 100.0,
        "safety": 50.0,
        "sycophancy": 33.3333,
        "gotchas": 26.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.7,
          "delta_mean": 0.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3478,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.6,
          "delta_mean": 0.5333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.4444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.48,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9714,
          "delta_mean": 0.1429,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7619,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4286,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 3,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.5266,
        "decode_tps_c4": 10.4745,
        "ttft_s": 1.0413,
        "prefill_tps_4k": 106.8456,
        "prefill_tps_12k": 94.8824,
        "load_seconds": 153.8431,
        "size_gb": 14.3339,
        "vram_gb": 14.3339,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-1-5b-instruct",
      "display_name": "Qwen2 1.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2:1.5b-instruct-q4_K_M",
      "size_bytes": 986000000,
      "model_version_id": 158,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 49.6912,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 6,
      "judge_a_mean": 4.18,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.5598,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.995,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 44.832,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 44.1,
      "ci_high": 55.3,
      "hard_score": 20.0768,
      "hard_n": 71,
      "ext_overall": 49.4115,
      "n_truncated": 14,
      "n_extended": 14,
      "n_gated": 5,
      "beats": [
        73,
        85,
        86,
        100,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        159,
        161,
        171,
        175,
        176,
        178,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 189,
      "class_rank": 23,
      "rank_lo": 175,
      "rank_hi": 203,
      "class_rank_lo": 18,
      "class_rank_hi": 31,
      "categories": {
        "reasoning": 28.5714,
        "math": 32.0,
        "coding": 30.6492,
        "instruction": 61.5556,
        "knowledge": 63.3333,
        "long_context": 27.7778,
        "multilingual": 48.3333,
        "tool_use": 80.0,
        "safety": 75.0,
        "sycophancy": 66.6667,
        "gotchas": 60.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3065,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8667,
          "delta_mean": 0.7667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6156,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8,
          "delta_mean": 1.9,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.32,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2571,
          "delta_mean": 0.6,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.25,
          "delta_mean": 0.9583,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 43.0686,
        "decode_tps_c4": 49.7986,
        "ttft_s": 0.118,
        "prefill_tps_4k": 1284.0797,
        "prefill_tps_12k": 1000.4801,
        "load_seconds": null,
        "size_gb": null,
        "vram_gb": null,
        "gpu_fraction": null,
        "context_length": null
      },
      "perf_units": {
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "ttft_s": "s"
      }
    },
    {
      "slug": "codeqwen-7b-chat-v1-5",
      "display_name": "Codeqwen 7B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "codeqwen:7b-chat-v1.5-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 77,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 49.3922,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.9152,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8069,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2265,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 3.3351,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 44.18,
      "ci_high": 54.71,
      "hard_score": 34.1895,
      "hard_n": 71,
      "ext_overall": 49.3922,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 8,
      "beats": [
        73,
        85,
        86,
        100,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 190,
      "class_rank": 65,
      "rank_lo": 176,
      "rank_hi": 204,
      "class_rank_lo": 62,
      "class_rank_hi": 67,
      "categories": {
        "reasoning": 25.7143,
        "math": 8.0,
        "coding": 69.3992,
        "instruction": 72.1111,
        "knowledge": 40.0,
        "long_context": 38.8889,
        "multilingual": 35.0,
        "tool_use": 95.0,
        "safety": 60.4167,
        "sycophancy": 66.6667,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.694,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.3,
          "delta_mean": 1.0667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7211,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8667,
          "delta_mean": 1.6333,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.8571,
          "delta_mean": 0.4,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6042,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2609,
          "delta_mean": 1.7826,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.95,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.75,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.8161,
        "decode_tps_c4": 25.2161,
        "ttft_s": 0.3825,
        "prefill_tps_4k": 315.6289,
        "prefill_tps_12k": 278.861,
        "load_seconds": 68.796,
        "size_gb": 13.6452,
        "vram_gb": 13.6452,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-r1-1-5b-qwen-distill",
      "display_name": "Deepseek R1 1.5B Qwen Distill",
      "family": "deepseek",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-r1:1.5b-qwen-distill-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 88,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 48.918,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 4,
      "judge_a_mean": 3.755,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.2238,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7945,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 44.2233,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 43.46,
      "ci_high": 54.73,
      "hard_score": 25.493,
      "hard_n": 71,
      "ext_overall": 53.149,
      "n_truncated": 60,
      "n_extended": 59,
      "n_gated": 8,
      "beats": [
        73,
        85,
        86,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        181,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 191,
      "class_rank": 24,
      "rank_lo": 177,
      "rank_hi": 205,
      "class_rank_lo": 19,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 71.4286,
        "math": 76.0,
        "coding": 24.4444,
        "instruction": 66.1667,
        "knowledge": 53.3333,
        "long_context": 22.2222,
        "multilingual": 47.5,
        "tool_use": 62.5,
        "safety": 16.6667,
        "sycophancy": 75.0,
        "gotchas": 73.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5333,
          "delta_mean": 0.3333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6617,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7,
          "delta_mean": 1.3333,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.76,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.475,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.7143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1714,
          "delta_mean": 0.6571,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.0476,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.7143,
          "delta_mean": 0.5238,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 3,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 44.8517,
        "decode_tps_c4": 50.5694,
        "ttft_s": 0.1019,
        "prefill_tps_4k": 1286.6169,
        "prefill_tps_12k": 1006.2381,
        "load_seconds": 26.9672,
        "size_gb": 5.0759,
        "vram_gb": 5.0759,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "hermes3-3b-llama3-2",
      "display_name": "Hermes3 3B Llama3.2",
      "family": "llama",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Llama 3.2",
      "license": "llama3",
      "quantization": "Q4_K_M",
      "ollama_ref": "hermes3:3b-llama3.2-q4_K_M",
      "size_bytes": 2000000000,
      "model_version_id": 128,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 48.4727,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.2533,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.0027,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.969,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 5.9772,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 42.53,
      "ci_high": 54.25,
      "hard_score": 14.8528,
      "hard_n": 71,
      "ext_overall": 48.4727,
      "n_truncated": 2,
      "n_extended": 1,
      "n_gated": 2,
      "beats": [
        73,
        85,
        86,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 192,
      "class_rank": 23,
      "rank_lo": 177,
      "rank_hi": 206,
      "class_rank_lo": 22,
      "class_rank_hi": 24,
      "categories": {
        "reasoning": 37.1429,
        "math": 28.0,
        "coding": 11.3333,
        "instruction": 76.4444,
        "knowledge": 63.3333,
        "long_context": 50.0,
        "multilingual": 43.3333,
        "tool_use": 60.0,
        "safety": 66.6667,
        "sycophancy": 83.3333,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1133,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7644,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5667,
          "delta_mean": 1.3667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.28,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8857,
          "delta_mean": 0.6286,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2609,
          "delta_mean": 1.0,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 23.3722,
        "decode_tps_c4": 50.7232,
        "ttft_s": 0.193,
        "prefill_tps_4k": 501.8384,
        "prefill_tps_12k": 375.8361,
        "load_seconds": 106.875,
        "size_gb": 17.4131,
        "vram_gb": 17.4131,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-moe-3b-instruct",
      "display_name": "Granite3 Moe 3B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3-moe:3b-instruct-q4_K_M",
      "size_bytes": 2100000000,
      "model_version_id": 121,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 47.72,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 4.7526,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8817,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2619,
      "agreement": "high",
      "disputed": 23,
      "tokens_per_s": 38.8409,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 42.33,
      "ci_high": 53.13,
      "hard_score": 20.2004,
      "hard_n": 71,
      "ext_overall": 47.4422,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 0,
      "beats": [
        73,
        85,
        86,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 193,
      "class_rank": 24,
      "rank_lo": 179,
      "rank_hi": 206,
      "class_rank_lo": 22,
      "class_rank_hi": 24,
      "categories": {
        "reasoning": 37.1429,
        "math": 20.0,
        "coding": 31.5595,
        "instruction": 67.1667,
        "knowledge": 76.6667,
        "long_context": 27.7778,
        "multilingual": 36.6667,
        "tool_use": 45.0,
        "safety": 87.5,
        "sycophancy": 50.0,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3156,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3,
          "delta_mean": 1.2333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6717,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.0333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.3,
          "delta_mean": 1.8,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6,
          "delta_mean": 0.9429,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.7083,
          "delta_mean": 1.3333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.45,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 147.4023,
        "decode_tps_c4": 218.43,
        "ttft_s": 0.0373,
        "prefill_tps_4k": 3846.3786,
        "prefill_tps_12k": 4339.7265,
        "load_seconds": 11.375,
        "size_gb": 4.2969,
        "vram_gb": 4.2969,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "smollm2-1-7b-instruct",
      "display_name": "Smollm2 1.7B Instruct",
      "family": "smollm",
      "org": null,
      "params_b": 1.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "smollm2:1.7b-instruct-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 174,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 47.0989,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.5833,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.793,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0983,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 9.2847,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 41.66,
      "ci_high": 52.7,
      "hard_score": 25.95,
      "hard_n": 71,
      "ext_overall": 47.0989,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 2,
      "beats": [
        73,
        85,
        86,
        108,
        109,
        120,
        145,
        148,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 194,
      "class_rank": 25,
      "rank_lo": 180,
      "rank_hi": 206,
      "class_rank_lo": 20,
      "class_rank_hi": 33,
      "categories": {
        "reasoning": 28.5714,
        "math": 8.0,
        "coding": 35.3191,
        "instruction": 79.5,
        "knowledge": 63.3333,
        "long_context": 33.3333,
        "multilingual": 35.0,
        "tool_use": 70.0,
        "safety": 70.8333,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3532,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9333,
          "delta_mean": 1.0333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.795,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.5,
          "delta_mean": 1.7333,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4286,
          "delta_mean": 0.6,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4783,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.625,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 41.7257,
        "decode_tps_c4": 53.2561,
        "ttft_s": 0.1093,
        "prefill_tps_4k": 670.9104,
        "prefill_tps_12k": 678.9333,
        "load_seconds": 95.0693,
        "size_gb": 14.2408,
        "vram_gb": 14.2408,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite4-350m-h",
      "display_name": "Granite4 350M H",
      "family": "granite",
      "org": null,
      "params_b": 0.35,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q8_0",
      "ollama_ref": "granite4:350m-h",
      "size_bytes": 366000000,
      "model_version_id": 124,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 45.5868,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 12,
      "judge_a_mean": 4.0379,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.1269,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2071,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 17.6163,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 40.67,
      "ci_high": 50.69,
      "hard_score": 25.1554,
      "hard_n": 71,
      "ext_overall": 45.5868,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 7,
      "beats": [
        73,
        85,
        86,
        108,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 195,
      "class_rank": 26,
      "rank_lo": 184,
      "rank_hi": 208,
      "class_rank_lo": 22,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 20.0,
        "math": 0.0,
        "coding": 36.0035,
        "instruction": 79.2778,
        "knowledge": 76.6667,
        "long_context": 16.6667,
        "multilingual": 43.3333,
        "tool_use": 80.0,
        "safety": 58.3333,
        "sycophancy": 41.6667,
        "gotchas": 20.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.36,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9667,
          "delta_mean": 1.3333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7928,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.8333,
          "disputed": 13,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.1143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.9429,
          "delta_mean": 0.2857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5833,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.8,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.375,
          "delta_mean": 1.75,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 55.7803,
        "decode_tps_c4": 128.2196,
        "ttft_s": 0.142,
        "prefill_tps_4k": 1739.4345,
        "prefill_tps_12k": 1756.292,
        "load_seconds": 8.367,
        "size_gb": 1.2489,
        "vram_gb": 1.2489,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "llama2-13b-chat",
      "display_name": "Llama2 13B Chat",
      "family": "llama",
      "org": null,
      "params_b": 13.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama2",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama2:13b-chat-q4_K_M",
      "size_bytes": 7900000000,
      "model_version_id": 133,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 45.302,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 4.5357,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8305,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9795,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 3.4621,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 40.18,
      "ci_high": 50.59,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 45.302,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 2,
      "beats": [
        73,
        85,
        86,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 196,
      "class_rank": 24,
      "rank_lo": 181,
      "rank_hi": 211,
      "class_rank_lo": 24,
      "class_rank_hi": 25,
      "categories": {
        "reasoning": 45.7143,
        "math": 8.0,
        "coding": 10.0589,
        "instruction": 47.8333,
        "knowledge": 73.3333,
        "long_context": 27.7778,
        "multilingual": 32.5,
        "tool_use": 75.0,
        "safety": 87.5,
        "sycophancy": 66.6667,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1006,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9667,
          "delta_mean": 0.7,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4783,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9,
          "delta_mean": 1.7333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.325,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2857,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.25,
          "delta_mean": 0.875,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.75,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 7.0661,
        "decode_tps_c4": 16.3759,
        "ttft_s": 0.6927,
        "prefill_tps_4k": 147.5167,
        "prefill_tps_12k": 159.6716,
        "load_seconds": 109.25,
        "size_gb": 21.457,
        "vram_gb": 21.457,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-1b-instruct",
      "display_name": "Falcon3 1B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3-tmplfix:1b-instruct-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 190,
      "run_id": "r20261028-bv1-7ce712",
      "battery_version": "v1",
      "tested_at": "2026-10-28T02:36:44.578Z",
      "overall": 44.3165,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.2981,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.7586,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2052,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 15.4134,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 39.76,
      "ci_high": 49.31,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 44.6471,
      "n_truncated": 16,
      "n_extended": 15,
      "n_gated": 0,
      "beats": [
        73,
        85,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 197,
      "class_rank": 27,
      "rank_lo": 180,
      "rank_hi": 212,
      "class_rank_lo": 19,
      "class_rank_hi": 36,
      "categories": {
        "reasoning": 40.0,
        "math": 20.0,
        "coding": 18.6815,
        "instruction": 64.8889,
        "knowledge": 70.0,
        "long_context": 11.1111,
        "multilingual": 34.1667,
        "tool_use": 40.0,
        "safety": 100.0,
        "sycophancy": 66.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1868,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5333,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6489,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2,
          "delta_mean": 1.7333,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1429,
          "delta_mean": 0.7429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 1.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.1667,
          "delta_mean": 1.5,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.75,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 51.4068,
        "decode_tps_c4": 107.3623,
        "ttft_s": 0.0801,
        "prefill_tps_4k": 1350.2017,
        "prefill_tps_12k": 1381.3319,
        "load_seconds": 35.6724,
        "size_gb": 1.057,
        "vram_gb": 1.057,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-7b-chat-v1-5",
      "display_name": "Qwen 7B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen:7b-chat-v1.5-q4_K_M",
      "size_bytes": 4800000000,
      "model_version_id": 156,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 44.091,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 4.38,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8057,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1048,
      "agreement": "high",
      "disputed": 20,
      "tokens_per_s": 4.7173,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 39.07,
      "ci_high": 48.93,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 44.091,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 1,
      "beats": [
        73,
        85,
        86,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 198,
      "class_rank": 66,
      "rank_lo": 183,
      "rank_hi": 211,
      "class_rank_lo": 65,
      "class_rank_hi": 68,
      "categories": {
        "reasoning": 45.7143,
        "math": 16.0,
        "coding": 5.855,
        "instruction": 71.8889,
        "knowledge": 76.6667,
        "long_context": 11.1111,
        "multilingual": 45.0,
        "tool_use": 35.0,
        "safety": 89.5833,
        "sycophancy": 75.0,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0585,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.5667,
          "delta_mean": 0.2333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7189,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5333,
          "delta_mean": 1.9333,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.45,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.4286,
          "delta_mean": 0.8571,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8958,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.625,
          "delta_mean": 1.375,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.75,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 1.125,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 12.3461,
        "decode_tps_c4": 23.1981,
        "ttft_s": 0.4216,
        "prefill_tps_4k": 249.4216,
        "prefill_tps_12k": 183.8565,
        "load_seconds": 256.75,
        "size_gb": 39.2427,
        "vram_gb": 39.2427,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "dolphin-phi-2-7b-v2-6",
      "display_name": "Dolphin Phi 2.7B V2.6",
      "family": "phi",
      "org": null,
      "params_b": 2.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Phi-2",
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "dolphin-phi:2.7b-v2.6-q4_K_M",
      "size_bytes": 1800000000,
      "model_version_id": 91,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 44.019,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 4.274,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.7137,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1174,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 6.7409,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 2048,
      "context_tokens": 2048,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 38.93,
      "ci_high": 49.35,
      "hard_score": 16.4405,
      "hard_n": 71,
      "ext_overall": 44.0005,
      "n_truncated": 21,
      "n_extended": 20,
      "n_gated": 1,
      "beats": [
        73,
        85,
        108,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 199,
      "class_rank": 28,
      "rank_lo": 185,
      "rank_hi": 209,
      "class_rank_lo": 22,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 31.4286,
        "math": 40.0,
        "coding": 28.5758,
        "instruction": 69.7778,
        "knowledge": 70.0,
        "long_context": 5.5556,
        "multilingual": 15.0,
        "tool_use": 65.0,
        "safety": 70.8333,
        "sycophancy": 58.3333,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2858,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3,
          "delta_mean": 0.6667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6978,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8667,
          "delta_mean": 1.6667,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6571,
          "delta_mean": 0.6286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8696,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 25.0211,
        "decode_tps_c4": 38.4384,
        "ttft_s": 0.1742,
        "prefill_tps_4k": 502.8911,
        "prefill_tps_12k": 515.4769,
        "load_seconds": 38.828,
        "size_gb": 7.1705,
        "vram_gb": 7.1705,
        "gpu_fraction": 1.0,
        "context_length": 2048.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen3-0-6b",
      "display_name": "Qwen3 0.6B",
      "family": "qwen",
      "org": null,
      "params_b": 0.6,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen3:0.6b-q4_K_M",
      "size_bytes": 523000000,
      "model_version_id": 165,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 43.9956,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.29,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.5029,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.121,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 216.8525,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 38.76,
      "ci_high": 49.44,
      "hard_score": 25.0383,
      "hard_n": 71,
      "ext_overall": 43.7244,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 8,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 200,
      "class_rank": 29,
      "rank_lo": 185,
      "rank_hi": 210,
      "class_rank_lo": 22,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 37.1429,
        "math": 40.0,
        "coding": 32.5394,
        "instruction": 74.3333,
        "knowledge": 20.0,
        "long_context": 11.1111,
        "multilingual": 48.3333,
        "tool_use": 82.5,
        "safety": 50.0,
        "sycophancy": 41.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3254,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.6667,
          "delta_mean": 1.0667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7433,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7667,
          "delta_mean": 1.8,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9143,
          "delta_mean": 0.5714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2917,
          "delta_mean": 1.2917,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.825,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 0.875,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 222.9919,
        "decode_tps_c4": 261.1199,
        "ttft_s": 0.0203,
        "prefill_tps_4k": 8611.2121,
        "prefill_tps_12k": 5013.0401,
        "load_seconds": 66.126,
        "size_gb": 15.6378,
        "vram_gb": 15.6378,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "meta-llama-llama-2-7b-chat-hf",
      "display_name": "Llama-2-7b-chat-hf",
      "family": "llama",
      "org": "meta-llama",
      "params_b": 6.74,
      "arch": "dense",
      "hf_downloads": 426308,
      "hf_rank": 358,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "llama2",
      "quantization": "Q4_K_M",
      "ollama_ref": "llama2:7b-chat-q4_K_M",
      "size_bytes": 4081978368,
      "model_version_id": 17,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": 43.7532,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.6233,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.0676,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.04,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 5.8499,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 8,
      "concurrency_text": "4 to 8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 39.0,
      "ci_high": 48.76,
      "hard_score": 16.4174,
      "hard_n": 38,
      "ext_overall": 43.7532,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 2,
      "beats": [
        73,
        85,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 201,
      "class_rank": 67,
      "rank_lo": 182,
      "rank_hi": 212,
      "class_rank_lo": 64,
      "class_rank_hi": 69,
      "categories": {
        "reasoning": 25.7143,
        "math": 4.0,
        "coding": 18.6197,
        "instruction": 72.1111,
        "knowledge": 66.6667,
        "long_context": 16.6667,
        "multilingual": 33.3333,
        "tool_use": 65.0,
        "safety": 91.6667,
        "sycophancy": 66.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1862,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6667,
          "delta_mean": 0.9667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7211,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2,
          "delta_mean": 1.2667,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.04,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9714,
          "delta_mean": 0.8,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.5,
          "delta_mean": 1.1667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 1.0,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.9642,
        "decode_tps_c4": 26.7272,
        "ttft_s": 0.4311,
        "prefill_tps_4k": 266.1857,
        "prefill_tps_12k": 291.4854,
        "load_seconds": 68.3621,
        "size_gb": 12.7905,
        "vram_gb": 12.7905,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "stablelm-zephyr-3b",
      "display_name": "Stablelm Zephyr 3B",
      "family": "stablelm",
      "org": null,
      "params_b": 2.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "StableLM 3B",
      "license": "stability-community",
      "quantization": "Q4_K_M",
      "ollama_ref": "stablelm-zephyr:3b-q4_K_M",
      "size_bytes": 1700000000,
      "model_version_id": 179,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 43.7042,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 12,
      "judge_a_mean": 4.479,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.8455,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9993,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 6.7269,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 38.41,
      "ci_high": 49.3,
      "hard_score": 19.9219,
      "hard_n": 71,
      "ext_overall": 43.7042,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 3,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186,
        193,
        202
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 202,
      "class_rank": 30,
      "rank_lo": 185,
      "rank_hi": 210,
      "class_rank_lo": 22,
      "class_rank_hi": 34,
      "categories": {
        "reasoning": 45.7143,
        "math": 16.0,
        "coding": 29.3458,
        "instruction": 60.0556,
        "knowledge": 50.0,
        "long_context": 22.2222,
        "multilingual": 34.1667,
        "tool_use": 65.0,
        "safety": 70.8333,
        "sycophancy": 50.0,
        "gotchas": 20.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2935,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4333,
          "delta_mean": 1.2333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6006,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0333,
          "delta_mean": 1.3667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4286,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8857,
          "delta_mean": 0.7714,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.0,
          "delta_mean": 0.875,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 26.539,
        "decode_tps_c4": 39.5994,
        "ttft_s": 0.2076,
        "prefill_tps_4k": 406.6915,
        "prefill_tps_12k": 523.5139,
        "load_seconds": 72.531,
        "size_gb": 12.5071,
        "vram_gb": 12.5071,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-1-moe-1b-instruct",
      "display_name": "Granite3.1 Moe 1B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 1.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3.1-moe:1b-instruct-q4_K_M",
      "size_bytes": 834000000,
      "model_version_id": 112,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 43.6252,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.3562,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.556,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0931,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 27.1371,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 37.87,
      "ci_high": 49.14,
      "hard_score": 18.9806,
      "hard_n": 71,
      "ext_overall": 44.2549,
      "n_truncated": 22,
      "n_extended": 22,
      "n_gated": 5,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 203,
      "class_rank": 31,
      "rank_lo": 185,
      "rank_hi": 212,
      "class_rank_lo": 22,
      "class_rank_hi": 35,
      "categories": {
        "reasoning": 31.4286,
        "math": 12.0,
        "coding": 38.1153,
        "instruction": 59.0,
        "knowledge": 63.3333,
        "long_context": 33.3333,
        "multilingual": 24.1667,
        "tool_use": 62.5,
        "safety": 68.75,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3812,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4667,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.59,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4333,
          "delta_mean": 1.9667,
          "disputed": 12,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.12,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1714,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2083,
          "delta_mean": 1.2083,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 102.6271,
        "decode_tps_c4": 149.4756,
        "ttft_s": 0.0649,
        "prefill_tps_4k": 2095.2439,
        "prefill_tps_12k": 1623.637,
        "load_seconds": 44.8459,
        "size_gb": 7.5097,
        "vram_gb": 7.5097,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-5-0-5b-instruct",
      "display_name": "Qwen2.5 0.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 0.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5:0.5b-instruct-q4_K_M",
      "size_bytes": 398000000,
      "model_version_id": 159,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 43.3114,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 3.7617,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.9155,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1424,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 98.1337,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 37.36,
      "ci_high": 49.36,
      "hard_score": 18.1676,
      "hard_n": 71,
      "ext_overall": 43.1769,
      "n_truncated": 12,
      "n_extended": 12,
      "n_gated": 9,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        157,
        161,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 204,
      "class_rank": 32,
      "rank_lo": 186,
      "rank_hi": 212,
      "class_rank_lo": 23,
      "class_rank_hi": 35,
      "categories": {
        "reasoning": 28.5714,
        "math": 36.0,
        "coding": 26.4537,
        "instruction": 54.8889,
        "knowledge": 56.6667,
        "long_context": 38.8889,
        "multilingual": 23.3333,
        "tool_use": 62.5,
        "safety": 62.5,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2645,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.1,
          "delta_mean": 0.6,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5489,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9,
          "delta_mean": 2.0333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3889,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.36,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2857,
          "delta_mean": 0.8286,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7917,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.625,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 99.2639,
        "decode_tps_c4": 116.8258,
        "ttft_s": 0.0475,
        "prefill_tps_4k": 3069.2425,
        "prefill_tps_12k": 2326.8192,
        "load_seconds": 13.6418,
        "size_gb": 2.2186,
        "vram_gb": 2.2186,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "stablelm2-12b-chat",
      "display_name": "Stablelm2 12B Chat",
      "family": "stablelm",
      "org": null,
      "params_b": 12.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "stability-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "stablelm2:12b-chat-q4_K_M",
      "size_bytes": 7400000000,
      "model_version_id": 181,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 42.4203,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 4.6676,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.106,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.866,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 1.1752,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 38.11,
      "ci_high": 46.72,
      "hard_score": 14.6518,
      "hard_n": 71,
      "ext_overall": 42.4203,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 205,
      "class_rank": 25,
      "rank_lo": 190,
      "rank_hi": 213,
      "class_rank_lo": 24,
      "class_rank_hi": 25,
      "categories": {
        "reasoning": 22.8571,
        "math": 8.0,
        "coding": 45.037,
        "instruction": 47.0,
        "knowledge": 70.0,
        "long_context": 5.5556,
        "multilingual": 20.0,
        "tool_use": 67.5,
        "safety": 95.8333,
        "sycophancy": 50.0,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4504,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.1667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4,
          "delta_mean": 0.8333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.47,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0667,
          "delta_mean": 1.4667,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.1714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7714,
          "delta_mean": 0.5714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.9583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 7.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 7.4167,
          "delta_mean": 0.9583,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.675,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.875,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 4.737,
        "decode_tps_c4": 6.7286,
        "ttft_s": 0.8462,
        "prefill_tps_4k": 84.0702,
        "prefill_tps_12k": 102.7095,
        "load_seconds": 50.187,
        "size_gb": 14.0397,
        "vram_gb": 14.0397,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "gemma-2b-instruct-v1-1",
      "display_name": "Gemma 2B Instruct V1.1",
      "family": "gemma",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "gemma:2b-instruct-v1.1-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 100,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 41.8888,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 11,
      "judge_a_mean": 4.3169,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.7674,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0805,
      "agreement": "high",
      "disputed": 17,
      "tokens_per_s": 11.2841,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 36.84,
      "ci_high": 47.47,
      "hard_score": 22.5478,
      "hard_n": 71,
      "ext_overall": 41.8888,
      "n_truncated": 8,
      "n_extended": 8,
      "n_gated": 4,
      "beats": [
        73,
        85,
        120,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 206,
      "class_rank": 33,
      "rank_lo": 188,
      "rank_hi": 213,
      "class_rank_lo": 23,
      "class_rank_hi": 36,
      "categories": {
        "reasoning": 25.7143,
        "math": 20.0,
        "coding": 25.2293,
        "instruction": 66.6111,
        "knowledge": 60.0,
        "long_context": 27.7778,
        "multilingual": 37.5,
        "tool_use": 35.0,
        "safety": 79.1667,
        "sycophancy": 50.0,
        "gotchas": 26.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2523,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.0,
          "delta_mean": 1.3667,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6661,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.5333,
          "delta_mean": 1.6,
          "disputed": 9,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.375,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4286,
          "delta_mean": 0.6857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.875,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0,
          "delta_mean": 0.5,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 29.2977,
        "decode_tps_c4": 43.4303,
        "ttft_s": 0.1584,
        "prefill_tps_4k": 860.6247,
        "prefill_tps_12k": 858.877,
        "load_seconds": 13.453,
        "size_gb": 3.3356,
        "vram_gb": 3.3356,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite-code-8b-instruct",
      "display_name": "Granite Code 8B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 8.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite-code:8b-instruct-q4_K_M",
      "size_bytes": 4900000000,
      "model_version_id": 109,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 40.6686,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 5,
      "judge_a_mean": 4.4098,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.6222,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0846,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 2.9061,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 35.34,
      "ci_high": 46.69,
      "hard_score": 20.6412,
      "hard_n": 71,
      "ext_overall": 40.8538,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 7,
      "beats": [
        73,
        85,
        145,
        152,
        153,
        155,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 207,
      "class_rank": 68,
      "rank_lo": 193,
      "rank_hi": 215,
      "class_rank_lo": 66,
      "class_rank_hi": 69,
      "categories": {
        "reasoning": 40.0,
        "math": 28.0,
        "coding": 55.7397,
        "instruction": 53.6667,
        "knowledge": 63.3333,
        "long_context": 27.7778,
        "multilingual": 41.6667,
        "tool_use": 10.0,
        "safety": 45.8333,
        "sycophancy": 58.3333,
        "gotchas": 66.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5574,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7667,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5367,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.3333,
          "delta_mean": 1.5333,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2778,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.28,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.1143,
          "delta_mean": 0.8571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5217,
          "delta_mean": 0.7826,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.375,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 10.4679,
        "decode_tps_c4": 15.0173,
        "ttft_s": 0.4725,
        "prefill_tps_4k": 221.6899,
        "prefill_tps_12k": 226.6381,
        "load_seconds": 41.422,
        "size_gb": 9.9646,
        "vram_gb": 9.9646,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "phi-2-7b-chat-v2",
      "display_name": "Phi 2.7B Chat V2",
      "family": "phi",
      "org": null,
      "params_b": 2.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi:2.7b-chat-v2-q4_K_M",
      "size_bytes": 1800000000,
      "model_version_id": 148,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 39.8792,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.0669,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.3875,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9969,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 6.3673,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 2048,
      "context_tokens": 2048,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 35.51,
      "ci_high": 44.18,
      "hard_score": 17.496,
      "hard_n": 71,
      "ext_overall": 40.3085,
      "n_truncated": 28,
      "n_extended": 28,
      "n_gated": 1,
      "beats": [
        73,
        85,
        145,
        152,
        153,
        155,
        161,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 208,
      "class_rank": 34,
      "rank_lo": 195,
      "rank_hi": 214,
      "class_rank_lo": 26,
      "class_rank_hi": 37,
      "categories": {
        "reasoning": 34.2857,
        "math": 0.0,
        "coding": 34.7102,
        "instruction": 60.3333,
        "knowledge": 83.3333,
        "long_context": 0.0,
        "multilingual": 18.3333,
        "tool_use": 42.5,
        "safety": 85.4167,
        "sycophancy": 33.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3471,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7667,
          "delta_mean": 0.6333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6033,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9333,
          "delta_mean": 1.5,
          "disputed": 6,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7429,
          "delta_mean": 0.5143,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.8542,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.8696,
          "delta_mean": 1.087,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.425,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.625,
          "delta_mean": 1.25,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 26.0076,
        "decode_tps_c4": 38.2132,
        "ttft_s": 0.1717,
        "prefill_tps_4k": 507.9095,
        "prefill_tps_12k": 528.6954,
        "load_seconds": 37.266,
        "size_gb": 7.1705,
        "vram_gb": 7.1705,
        "gpu_fraction": 1.0,
        "context_length": 2048.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite-code-3b-instruct",
      "display_name": "Granite Code 3B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite-code:3b-instruct-q4_K_M",
      "size_bytes": 2100000000,
      "model_version_id": 108,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 37.9728,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 4.1136,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.2259,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2137,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 22.6418,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 2048,
      "context_tokens": 2048,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 32.65,
      "ci_high": 43.45,
      "hard_score": 25.598,
      "hard_n": 71,
      "ext_overall": 37.6025,
      "n_truncated": 10,
      "n_extended": 10,
      "n_gated": 8,
      "beats": [
        73,
        85,
        152,
        153,
        155,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 209,
      "class_rank": 25,
      "rank_lo": 197,
      "rank_hi": 216,
      "class_rank_lo": 25,
      "class_rank_hi": 25,
      "categories": {
        "reasoning": 37.1429,
        "math": 24.0,
        "coding": 45.557,
        "instruction": 54.2222,
        "knowledge": 56.6667,
        "long_context": 16.6667,
        "multilingual": 42.5,
        "tool_use": 15.0,
        "safety": 50.0,
        "sycophancy": 58.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9667,
          "delta_mean": 1.0667,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5422,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5333,
          "delta_mean": 1.5333,
          "disputed": 5,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.24,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.425,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.4286,
          "delta_mean": 0.6857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.8261,
          "delta_mean": 0.7826,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.375,
          "delta_mean": 2.0,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 71.8242,
        "decode_tps_c4": 107.2588,
        "ttft_s": 0.0658,
        "prefill_tps_4k": 1511.3038,
        "prefill_tps_12k": 1715.0365,
        "load_seconds": 25.438,
        "size_gb": 7.6715,
        "vram_gb": 7.6715,
        "gpu_fraction": 1.0,
        "context_length": 2048.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-0-5b-instruct",
      "display_name": "Qwen2 0.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 0.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2:0.5b-instruct-q4_K_M",
      "size_bytes": 398000000,
      "model_version_id": 157,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 36.0142,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 11,
      "judge_a_mean": 3.1543,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.3652,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0638,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 32.4191,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 30.93,
      "ci_high": 41.21,
      "hard_score": 15.6007,
      "hard_n": 71,
      "ext_overall": 36.9613,
      "n_truncated": 22,
      "n_extended": 21,
      "n_gated": 8,
      "beats": [
        73,
        85,
        152,
        153,
        155,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 210,
      "class_rank": 35,
      "rank_lo": 201,
      "rank_hi": 216,
      "class_rank_lo": 32,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 31.4286,
        "math": 8.0,
        "coding": 9.8106,
        "instruction": 50.1667,
        "knowledge": 43.3333,
        "long_context": 22.2222,
        "multilingual": 23.3333,
        "tool_use": 77.5,
        "safety": 58.3333,
        "sycophancy": 41.6667,
        "gotchas": 26.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0981,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.8667,
          "delta_mean": 0.6333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5017,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.2333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2333,
          "delta_mean": 2.0,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.2222,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.3714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1429,
          "delta_mean": 0.6857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.4583,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9583,
          "delta_mean": 0.75,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.625,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 93.7428,
        "decode_tps_c4": 137.5347,
        "ttft_s": 0.0505,
        "prefill_tps_4k": 3110.1724,
        "prefill_tps_12k": 2336.2753,
        "load_seconds": 14.0775,
        "size_gb": 2.2186,
        "vram_gb": 2.2186,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "granite3-moe-1b-instruct",
      "display_name": "Granite3 Moe 1B Instruct",
      "family": "granite",
      "org": null,
      "params_b": 1.0,
      "arch": "moe",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "granite3-moe:1b-instruct-q4_K_M",
      "size_bytes": 822000000,
      "model_version_id": 120,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 35.5124,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 3.7617,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.105,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9724,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 66.8333,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 30.96,
      "ci_high": 40.5,
      "hard_score": 18.6907,
      "hard_n": 71,
      "ext_overall": 35.3272,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 4,
      "beats": [
        73,
        85,
        152,
        153,
        171,
        175,
        176,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 211,
      "class_rank": 36,
      "rank_lo": 203,
      "rank_hi": 217,
      "class_rank_lo": 33,
      "class_rank_hi": 39,
      "categories": {
        "reasoning": 17.1429,
        "math": 8.0,
        "coding": 27.6354,
        "instruction": 59.0556,
        "knowledge": 36.6667,
        "long_context": 11.1111,
        "multilingual": 20.8333,
        "tool_use": 60.0,
        "safety": 79.1667,
        "sycophancy": 58.3333,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2764,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.8,
          "delta_mean": 1.3,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5906,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.5667,
          "delta_mean": 1.4667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.1714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.4,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2,
          "delta_mean": 0.4286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.7917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.9167,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.7083,
          "delta_mean": 1.2917,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.25,
          "delta_mean": 0.375,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 69.9666,
        "decode_tps_c4": 82.3603,
        "ttft_s": 0.0791,
        "prefill_tps_4k": 1882.2574,
        "prefill_tps_12k": 2179.6198,
        "load_seconds": 16.6659,
        "size_gb": 2.5995,
        "vram_gb": 2.3278,
        "gpu_fraction": 0.8955,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-coder-6-7b-instruct",
      "display_name": "Deepseek Coder 6.7B Instruct",
      "family": "deepseek",
      "org": null,
      "params_b": 6.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-coder:6.7b-instruct-q4_K_M",
      "size_bytes": 4100000000,
      "model_version_id": 86,
      "run_id": "r20261022-bv1-54a19b",
      "battery_version": "v1",
      "tested_at": "2026-10-23T09:25:14.997Z",
      "overall": 34.6862,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 3.2379,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.504,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8219,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 5.172,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 4,
      "concurrency_max": 4,
      "concurrency_text": "4 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 4 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 29.67,
      "ci_high": 39.96,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 34.6862,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 16,
      "beats": [
        73,
        85,
        152,
        153,
        171,
        175,
        178,
        180
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 212,
      "class_rank": 69,
      "rank_lo": 198,
      "rank_hi": 219,
      "class_rank_lo": 67,
      "class_rank_hi": 69,
      "categories": {
        "reasoning": 8.5714,
        "math": 0.0,
        "coding": 68.5486,
        "instruction": 53.9444,
        "knowledge": 10.0,
        "long_context": 61.1111,
        "multilingual": 21.6667,
        "tool_use": 55.0,
        "safety": 33.3333,
        "sycophancy": 50.0,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6855,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 6.1,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.2,
          "delta_mean": 1.0333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5394,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4,
          "delta_mean": 1.3333,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.6111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6286,
          "delta_mean": 0.7429,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2917,
          "delta_mean": 0.75,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.55,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.0,
          "delta_mean": 0.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 13.5797,
        "decode_tps_c4": 22.5,
        "ttft_s": 0.4079,
        "prefill_tps_4k": 277.2862,
        "prefill_tps_12k": 196.6318,
        "load_seconds": 262.032,
        "size_gb": 38.8391,
        "vram_gb": 38.8391,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-coder-33b-instruct",
      "display_name": "Deepseek Coder 33B Instruct",
      "family": "deepseek",
      "org": null,
      "params_b": 33.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-coder:33b-instruct-q4_K_M",
      "size_bytes": 19941000000,
      "model_version_id": 202,
      "run_id": "r20261102-bv1-5246ab",
      "battery_version": "v1",
      "tested_at": "2026-11-05T09:06:45.362Z",
      "overall": 33.4378,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 7,
      "judge_a_mean": 3.25,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.8119,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7943,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 0.8689,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "16-35",
      "size_class_label": "16–35B",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 28.95,
      "ci_high": 38.17,
      "hard_score": 30.9709,
      "hard_n": 35,
      "ext_overall": 33.4378,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 14,
      "beats": [
        73,
        85,
        152,
        171,
        175,
        178
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 213,
      "class_rank": 49,
      "rank_lo": 203,
      "rank_hi": 221,
      "class_rank_lo": 49,
      "class_rank_hi": 49,
      "categories": {
        "reasoning": 31.4286,
        "math": 12.0,
        "coding": 67.1786,
        "instruction": 46.1667,
        "knowledge": 6.6667,
        "long_context": 0.0,
        "multilingual": 25.8333,
        "tool_use": 70.0,
        "safety": 41.6667,
        "sycophancy": 58.3333,
        "gotchas": 53.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6718,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.1667,
          "delta_mean": 1.2333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4617,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6667,
          "delta_mean": 1.3333,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.12,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.3143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.1429,
          "delta_mean": 0.5714,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7083,
          "delta_mean": 0.5833,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.375,
          "delta_mean": 0.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 3.0045,
        "decode_tps_c4": 5.93,
        "ttft_s": 1.6182,
        "prefill_tps_4k": 63.6249,
        "prefill_tps_12k": 54.3164,
        "load_seconds": 239.9373,
        "size_gb": 53.4823,
        "vram_gb": 53.4823,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-math-1-5b-instruct",
      "display_name": "Qwen2 Math 1.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2-math-tmplfix:1.5b-instruct-q4_K_M",
      "size_bytes": 986000000,
      "model_version_id": 193,
      "run_id": "r20261028-bv1-ef0895",
      "battery_version": "v1",
      "tested_at": "2026-10-28T10:42:23.529Z",
      "overall": 33.1358,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 2.9864,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.2598,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0605,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 14.5632,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": null,
      "ci_low": 28.04,
      "ci_high": 38.22,
      "hard_score": null,
      "hard_n": 0,
      "ext_overall": 33.018,
      "n_truncated": 21,
      "n_extended": 21,
      "n_gated": 10,
      "beats": [
        73,
        85,
        152,
        175,
        178
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": false,
      "is_latest": true,
      "rank": 214,
      "class_rank": 37,
      "rank_lo": 203,
      "rank_hi": 222,
      "class_rank_lo": 31,
      "class_rank_hi": 44,
      "categories": {
        "reasoning": 65.7143,
        "math": 72.0,
        "coding": 17.4524,
        "instruction": 14.4444,
        "knowledge": 33.3333,
        "long_context": 11.1111,
        "multilingual": 37.5,
        "tool_use": 17.5,
        "safety": 29.1667,
        "sycophancy": 83.3333,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1745,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.0,
          "delta_mean": 0.6333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1444,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.7,
          "delta_mean": 1.0333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.72,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.375,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.6571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 6.0571,
          "delta_mean": 0.8857,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.2917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4167,
          "delta_mean": 0.875,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.8333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.175,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.125,
          "delta_mean": 1.875,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 45.2311,
        "decode_tps_c4": 62.0286,
        "ttft_s": 0.103,
        "prefill_tps_4k": 1216.6652,
        "prefill_tps_12k": 1324.363,
        "load_seconds": 11.5905,
        "size_gb": 0.986,
        "vram_gb": 0.986,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-5-coder-0-5b-instruct",
      "display_name": "Qwen2.5 Coder 0.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 0.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2.5-coder:0.5b-instruct-q4_K_M",
      "size_bytes": 398000000,
      "model_version_id": 161,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 32.8247,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 3.4214,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.3602,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1394,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 33.0638,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 27.52,
      "ci_high": 38.15,
      "hard_score": 21.9587,
      "hard_n": 71,
      "ext_overall": 33.0716,
      "n_truncated": 16,
      "n_extended": 14,
      "n_gated": 12,
      "beats": [
        73,
        85,
        152,
        153,
        171,
        175,
        178,
        180,
        186
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 215,
      "class_rank": 38,
      "rank_lo": 208,
      "rank_hi": 218,
      "class_rank_lo": 35,
      "class_rank_hi": 40,
      "categories": {
        "reasoning": 22.8571,
        "math": 8.0,
        "coding": 35.8981,
        "instruction": 49.5,
        "knowledge": 20.0,
        "long_context": 33.3333,
        "multilingual": 31.6667,
        "tool_use": 65.0,
        "safety": 29.1667,
        "sycophancy": 50.0,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.359,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.9,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0,
          "delta_mean": 1.0333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.495,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3333,
          "delta_mean": 1.8,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.6571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.1143,
          "delta_mean": 0.6571,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.2917,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4783,
          "delta_mean": 0.9565,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.65,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.875,
          "delta_mean": 1.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 96.6871,
        "decode_tps_c4": 172.544,
        "ttft_s": 0.0453,
        "prefill_tps_4k": 3051.7329,
        "prefill_tps_12k": 2335.6154,
        "load_seconds": 13.6408,
        "size_gb": 2.2186,
        "vram_gb": 2.2186,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "opencoder-1-5b-instruct",
      "display_name": "Opencoder 1.5B Instruct",
      "family": "other",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "opencoder:1.5b-instruct-q4_K_M",
      "size_bytes": 1400000000,
      "model_version_id": 145,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 31.9417,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 11,
      "judge_a_mean": 3.4095,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.2819,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2172,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 15.1339,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 27.6,
      "ci_high": 36.39,
      "hard_score": 21.5311,
      "hard_n": 71,
      "ext_overall": 31.9417,
      "n_truncated": 24,
      "n_extended": 22,
      "n_gated": 9,
      "beats": [
        73,
        85,
        152,
        153,
        171,
        175,
        178,
        180
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 216,
      "class_rank": 39,
      "rank_lo": 209,
      "rank_hi": 219,
      "class_rank_lo": 35,
      "class_rank_hi": 41,
      "categories": {
        "reasoning": 2.8571,
        "math": 0.0,
        "coding": 57.313,
        "instruction": 50.7778,
        "knowledge": 20.0,
        "long_context": 11.1111,
        "multilingual": 24.1667,
        "tool_use": 77.5,
        "safety": 43.75,
        "sycophancy": 58.3333,
        "gotchas": 26.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5731,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.0667,
          "delta_mean": 0.6333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5078,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.3,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.4333,
          "delta_mean": 1.9333,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2417,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.4857,
          "delta_mean": 0.4,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4375,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.1739,
          "delta_mean": 0.8696,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5833,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.775,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.25,
          "delta_mean": 2.25,
          "disputed": 4,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 17.4197,
        "decode_tps_c4": 16.2125,
        "ttft_s": 0.2167,
        "prefill_tps_4k": 361.0939,
        "prefill_tps_12k": 495.7045,
        "load_seconds": 47.3554,
        "size_gb": 8.4254,
        "vram_gb": 8.4254,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-4b-chat-v1-5",
      "display_name": "Qwen 4B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 4.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "tongyi-qianwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen:4b-chat-v1.5-q4_K_M",
      "size_bytes": 2500000000,
      "model_version_id": 155,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 30.1605,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 12,
      "judge_a_mean": 3.151,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.2902,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9721,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 56.8895,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 25.96,
      "ci_high": 34.53,
      "hard_score": 15.1006,
      "hard_n": 71,
      "ext_overall": 30.0679,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 6,
      "beats": [
        73,
        85,
        152,
        153,
        171,
        175,
        178,
        180
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 217,
      "class_rank": 26,
      "rank_lo": 211,
      "rank_hi": 219,
      "class_rank_lo": 26,
      "class_rank_hi": 26,
      "categories": {
        "reasoning": 11.4286,
        "math": 4.0,
        "coding": 3.0714,
        "instruction": 60.4444,
        "knowledge": 73.3333,
        "long_context": 0.0,
        "multilingual": 37.5,
        "tool_use": 15.0,
        "safety": 66.6667,
        "sycophancy": 50.0,
        "gotchas": 20.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0307,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.2667,
          "delta_mean": 0.1,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6044,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6333,
          "delta_mean": 1.5667,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.04,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.375,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.1143,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.7714,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3429,
          "delta_mean": 0.4857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 4.2083,
          "delta_mean": 1.0833,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.0,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 67.4728,
        "decode_tps_c4": 72.1522,
        "ttft_s": 0.0675,
        "prefill_tps_4k": 1750.3491,
        "prefill_tps_12k": 1243.2708,
        "load_seconds": 249.203,
        "size_gb": 56.0779,
        "vram_gb": 56.0779,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "smollm2-360m-instruct",
      "display_name": "Smollm2 360M Instruct",
      "family": "smollm",
      "org": null,
      "params_b": 0.36,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "smollm2:360m-instruct-q4_K_M",
      "size_bytes": 271000000,
      "model_version_id": 176,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 28.0758,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 8,
      "judge_a_mean": 3.3064,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.5453,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9766,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 173.9401,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 23.4,
      "ci_high": 32.93,
      "hard_score": 14.1409,
      "hard_n": 71,
      "ext_overall": 28.1506,
      "n_truncated": 26,
      "n_extended": 23,
      "n_gated": 7,
      "beats": [
        73,
        152,
        175,
        178
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 218,
      "class_rank": 40,
      "rank_lo": 212,
      "rank_hi": 223,
      "class_rank_lo": 37,
      "class_rank_hi": 45,
      "categories": {
        "reasoning": 8.5714,
        "math": 8.0,
        "coding": 23.8052,
        "instruction": 59.9444,
        "knowledge": 46.6667,
        "long_context": 11.1111,
        "multilingual": 21.6667,
        "tool_use": 25.0,
        "safety": 47.9167,
        "sycophancy": 41.6667,
        "gotchas": 46.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2381,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.0,
          "delta_mean": 0.8667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5994,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.2,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.9,
          "delta_mean": 1.3667,
          "disputed": 7,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.08,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.4571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3429,
          "delta_mean": 0.2857,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4792,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.6087,
          "delta_mean": 0.7391,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.875,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 275.3464,
        "decode_tps_c4": 91.1657,
        "ttft_s": 0.0152,
        "prefill_tps_4k": 9527.2994,
        "prefill_tps_12k": 2732.1585,
        "load_seconds": 13.6042,
        "size_gb": 3.0812,
        "vram_gb": 3.0812,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "yi-coder-1-5b-chat",
      "display_name": "YI Coder 1.5B Chat",
      "family": "yi",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "yi-coder:1.5b-chat-q4_K_M",
      "size_bytes": 964000000,
      "model_version_id": 186,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 26.5009,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 10,
      "judge_a_mean": 3.2883,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.2713,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1351,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 12.1315,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 22.13,
      "ci_high": 31.51,
      "hard_score": 21.6609,
      "hard_n": 71,
      "ext_overall": 26.4083,
      "n_truncated": 27,
      "n_extended": 27,
      "n_gated": 13,
      "beats": [
        73,
        152,
        175
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 219,
      "class_rank": 41,
      "rank_lo": 213,
      "rank_hi": 224,
      "class_rank_lo": 38,
      "class_rank_hi": 46,
      "categories": {
        "reasoning": 17.1429,
        "math": 0.0,
        "coding": 57.6708,
        "instruction": 36.0556,
        "knowledge": 53.3333,
        "long_context": 5.5556,
        "multilingual": 10.8333,
        "tool_use": 35.0,
        "safety": 22.9167,
        "sycophancy": 41.6667,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5767,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7667,
          "delta_mean": 0.9,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3606,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.1333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6333,
          "delta_mean": 1.5667,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1083,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.1714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.1143,
          "delta_mean": 0.7143,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.2292,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2174,
          "delta_mean": 0.8696,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.35,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.625,
          "delta_mean": 1.625,
          "disputed": 0,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 49.4729,
        "decode_tps_c4": 88.7185,
        "ttft_s": 0.1006,
        "prefill_tps_4k": 1051.3441,
        "prefill_tps_12k": 682.2714,
        "load_seconds": 183.9436,
        "size_gb": 26.9278,
        "vram_gb": 26.9278,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-1-8b-chat-v1-5",
      "display_name": "Qwen 1.8B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 1.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "tongyi-qianwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen:1.8b-chat-v1.5-q4_K_M",
      "size_bytes": 1200000000,
      "model_version_id": 153,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 24.6994,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 2.5269,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.9653,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.795,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 41.1793,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 20.23,
      "ci_high": 29.29,
      "hard_score": 12.6009,
      "hard_n": 71,
      "ext_overall": 24.6994,
      "n_truncated": 43,
      "n_extended": 42,
      "n_gated": 8,
      "beats": [
        73,
        152
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 220,
      "class_rank": 42,
      "rank_lo": 216,
      "rank_hi": 225,
      "class_rank_lo": 39,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 20.0,
        "math": 4.0,
        "coding": 2.35,
        "instruction": 38.4444,
        "knowledge": 60.0,
        "long_context": 0.0,
        "multilingual": 27.5,
        "tool_use": 20.0,
        "safety": 50.0,
        "sycophancy": 66.6667,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0235,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.3333,
          "delta_mean": 0.2,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3844,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.6667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.3,
          "delta_mean": 1.3667,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.04,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.275,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.3429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2857,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0417,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.7826,
          "delta_mean": 0.8261,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.25,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.125,
          "delta_mean": 1.125,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 127.3801,
        "decode_tps_c4": 82.5127,
        "ttft_s": 0.0307,
        "prefill_tps_4k": 3774.0783,
        "prefill_tps_12k": 2310.5373,
        "load_seconds": 115.8952,
        "size_gb": 27.0761,
        "vram_gb": 27.0761,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "stablelm2-1-6b-chat",
      "display_name": "Stablelm2 1.6B Chat",
      "family": "stablelm",
      "org": null,
      "params_b": 1.6,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "stability-nc",
      "quantization": "Q4_K_M",
      "ollama_ref": "stablelm2:1.6b-chat-q4_K_M",
      "size_bytes": 1000000000,
      "model_version_id": 180,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 24.3833,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 13,
      "judge_a_mean": 2.8631,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.3091,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9004,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 13.0479,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 20.24,
      "ci_high": 29.0,
      "hard_score": 10.5867,
      "hard_n": 71,
      "ext_overall": 24.1981,
      "n_truncated": 5,
      "n_extended": 4,
      "n_gated": 6,
      "beats": [
        73,
        152
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 221,
      "class_rank": 43,
      "rank_lo": 216,
      "rank_hi": 225,
      "class_rank_lo": 39,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 8.5714,
        "math": 4.0,
        "coding": 26.1006,
        "instruction": 31.6111,
        "knowledge": 40.0,
        "long_context": 0.0,
        "multilingual": 16.6667,
        "tool_use": 42.5,
        "safety": 50.0,
        "sycophancy": 41.6667,
        "gotchas": 13.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.261,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.4333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.9333,
          "delta_mean": 0.7667,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.1333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3161,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6333,
          "delta_mean": 1.4333,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.04,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5714,
          "delta_mean": 0.4,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.7826,
          "delta_mean": 0.6522,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.425,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.625,
          "delta_mean": 1.25,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 47.8704,
        "decode_tps_c4": 62.2439,
        "ttft_s": 0.0986,
        "prefill_tps_4k": 793.7922,
        "prefill_tps_12k": 981.995,
        "load_seconds": 44.6845,
        "size_gb": 7.4659,
        "vram_gb": 7.4659,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "sailor2-1b-chat",
      "display_name": "Sailor2 1B Chat",
      "family": "other",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": "Qwen 2.5",
      "license": null,
      "quantization": "Q4_K_M",
      "ollama_ref": "sailor2:1b-chat-q4_K_M",
      "size_bytes": 739000000,
      "model_version_id": 171,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 23.4625,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 14,
      "judge_a_mean": 2.6238,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.8657,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9818,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 18.2847,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 18.95,
      "ci_high": 28.09,
      "hard_score": 6.3866,
      "hard_n": 71,
      "ext_overall": 23.6477,
      "n_truncated": 14,
      "n_extended": 12,
      "n_gated": 10,
      "beats": [
        73,
        152
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 222,
      "class_rank": 44,
      "rank_lo": 217,
      "rank_hi": 225,
      "class_rank_lo": 39,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 8.5714,
        "math": 0.0,
        "coding": 10.5355,
        "instruction": 29.5556,
        "knowledge": 60.0,
        "long_context": 16.6667,
        "multilingual": 25.0,
        "tool_use": 27.5,
        "safety": 33.3333,
        "sycophancy": 66.6667,
        "gotchas": 6.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1054,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.9333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.0333,
          "delta_mean": 1.0333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.0667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.2956,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.9667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5333,
          "delta_mean": 1.4333,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.6,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0857,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5714,
          "delta_mean": 0.4857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.7083,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5652,
          "delta_mean": 0.9565,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.6667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.275,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.625,
          "delta_mean": 1.0,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 57.0506,
        "decode_tps_c4": 116.3405,
        "ttft_s": 0.0855,
        "prefill_tps_4k": 1537.8271,
        "prefill_tps_12k": 1148.714,
        "load_seconds": 23.3432,
        "size_gb": 4.0764,
        "vram_gb": 4.0764,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "stable-code-3b-instruct",
      "display_name": "Stable Code 3B Instruct",
      "family": "other",
      "org": null,
      "params_b": 2.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "stability-community",
      "quantization": "Q4_K_M",
      "ollama_ref": "stable-code:3b-instruct-q4_K_M",
      "size_bytes": 1700000000,
      "model_version_id": 178,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 22.808,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 9,
      "judge_a_mean": 2.9271,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.0723,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0196,
      "agreement": "high",
      "disputed": 10,
      "tokens_per_s": 4.8021,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 18.4,
      "ci_high": 27.4,
      "hard_score": 17.3937,
      "hard_n": 71,
      "ext_overall": 22.808,
      "n_truncated": 31,
      "n_extended": 29,
      "n_gated": 10,
      "beats": [
        73,
        152
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 223,
      "class_rank": 45,
      "rank_lo": 219,
      "rank_hi": 225,
      "class_rank_lo": 41,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 22.8571,
        "math": 0.0,
        "coding": 6.4983,
        "instruction": 55.7778,
        "knowledge": 33.3333,
        "long_context": 5.5556,
        "multilingual": 26.6667,
        "tool_use": 15.0,
        "safety": 39.5833,
        "sycophancy": 41.6667,
        "gotchas": 40.0,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.065,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.7667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.6667,
          "delta_mean": 0.2333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5578,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.7333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.0667,
          "delta_mean": 1.6667,
          "disputed": 5,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8857,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2857,
          "delta_mean": 0.8286,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.3958,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.2174,
          "delta_mean": 0.8696,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.4167,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.125,
          "delta_mean": 1.5,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 26.4371,
        "decode_tps_c4": 38.6196,
        "ttft_s": 0.2128,
        "prefill_tps_4k": 403.7822,
        "prefill_tps_12k": 224.2395,
        "load_seconds": 316.109,
        "size_gb": 44.905,
        "vram_gb": 44.905,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "deepseek-coder-1-3b-instruct",
      "display_name": "Deepseek Coder 1.3B Instruct",
      "family": "deepseek",
      "org": null,
      "params_b": 1.3,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "deepseek-coder:1.3b-instruct-q4_K_M",
      "size_bytes": 874000000,
      "model_version_id": 85,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 22.6117,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 14,
      "judge_a_mean": 2.4029,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.6605,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8224,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 51.9358,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 18.27,
      "ci_high": 27.17,
      "hard_score": 21.8899,
      "hard_n": 71,
      "ext_overall": 22.2942,
      "n_truncated": 7,
      "n_extended": 7,
      "n_gated": 16,
      "beats": [
        73,
        152
      ],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 224,
      "class_rank": 46,
      "rank_lo": 218,
      "rank_hi": 225,
      "class_rank_lo": 40,
      "class_rank_hi": 47,
      "categories": {
        "reasoning": 5.7143,
        "math": 0.0,
        "coding": 54.5132,
        "instruction": 40.5,
        "knowledge": 3.3333,
        "long_context": 11.1111,
        "multilingual": 23.3333,
        "tool_use": 40.0,
        "safety": 25.0,
        "sycophancy": 33.3333,
        "gotchas": 6.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5451,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 3.6667,
          "delta_mean": 0.9333,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.0667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.405,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.6,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9667,
          "delta_mean": 1.6333,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.1111,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0571,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.7143,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.0857,
          "delta_mean": 0.6286,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.2917,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.9583,
          "delta_mean": 0.6667,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.4,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.625,
          "delta_mean": 0.25,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 53.6227,
        "decode_tps_c4": 64.1043,
        "ttft_s": 0.1016,
        "prefill_tps_4k": 1056.5058,
        "prefill_tps_12k": 687.7793,
        "load_seconds": 182.5082,
        "size_gb": 26.8745,
        "vram_gb": 26.8745,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "smollm2-135m-instruct",
      "display_name": "Smollm2 135M Instruct",
      "family": "smollm",
      "org": null,
      "params_b": 0.135,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "smollm2:135m-instruct-q4_K_M",
      "size_bytes": 105000000,
      "model_version_id": 175,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 20.1171,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 14,
      "judge_a_mean": 2.3052,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.3941,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0694,
      "agreement": "high",
      "disputed": 11,
      "tokens_per_s": 177.3219,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 16.19,
      "ci_high": 24.59,
      "hard_score": 11.2153,
      "hard_n": 71,
      "ext_overall": 20.4907,
      "n_truncated": 55,
      "n_extended": 52,
      "n_gated": 9,
      "beats": [],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 225,
      "class_rank": 47,
      "rank_lo": 220,
      "rank_hi": 227,
      "class_rank_lo": 42,
      "class_rank_hi": 49,
      "categories": {
        "reasoning": 2.8571,
        "math": 0.0,
        "coding": 14.0857,
        "instruction": 36.0556,
        "knowledge": 36.6667,
        "long_context": 5.5556,
        "multilingual": 15.0,
        "tool_use": 25.0,
        "safety": 45.8333,
        "sycophancy": 50.0,
        "gotchas": 6.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1409,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.1667,
          "delta_mean": 0.7333,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.0667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3606,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3,
          "delta_mean": 1.7,
          "disputed": 7,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.3667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0556,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.0286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.1429,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.7429,
          "delta_mean": 0.4571,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4583,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.75,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2609,
          "delta_mean": 0.9565,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.5,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.25,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5,
          "delta_mean": 1.5,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 179.8502,
        "decode_tps_c4": 201.1696,
        "ttft_s": 0.025,
        "prefill_tps_4k": 4627.7033,
        "prefill_tps_12k": 4923.8474,
        "load_seconds": 16.205,
        "size_gb": 1.7039,
        "vram_gb": 1.7039,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen-0-5b-chat-v1-5",
      "display_name": "Qwen 0.5B Chat V1.5",
      "family": "qwen",
      "org": null,
      "params_b": 0.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "tongyi-qianwen-research",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen:0.5b-chat-v1.5-q4_K_M",
      "size_bytes": 407000000,
      "model_version_id": 152,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 16.1421,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 11,
      "judge_a_mean": 2.6067,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.5023,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.2606,
      "agreement": "high",
      "disputed": 15,
      "tokens_per_s": 179.503,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": 12.51,
      "ci_high": 20.27,
      "hard_score": 15.234,
      "hard_n": 71,
      "ext_overall": 16.1421,
      "n_truncated": 10,
      "n_extended": 8,
      "n_gated": 16,
      "beats": [],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 226,
      "class_rank": 48,
      "rank_lo": 225,
      "rank_hi": 227,
      "class_rank_lo": 47,
      "class_rank_hi": 49,
      "categories": {
        "reasoning": 14.2857,
        "math": 0.0,
        "coding": 8.9654,
        "instruction": 40.7778,
        "knowledge": 13.3333,
        "long_context": 0.0,
        "multilingual": 30.0,
        "tool_use": 15.0,
        "safety": 22.9167,
        "sycophancy": 33.3333,
        "gotchas": 26.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0897,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.5667,
          "delta_mean": 0.4333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.2667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.4078,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.5333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.6333,
          "delta_mean": 1.9667,
          "disputed": 11,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.3,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.1429,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.3714,
          "delta_mean": 0.6857,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.2292,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.375,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.5652,
          "delta_mean": 1.2174,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.15,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 3.125,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.375,
          "delta_mean": 2.0,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 99.844,
        "decode_tps_c4": 98.9439,
        "ttft_s": 0.054,
        "prefill_tps_4k": 1951.2259,
        "prefill_tps_12k": 1103.804,
        "load_seconds": 105.1233,
        "size_gb": 13.3685,
        "vram_gb": 13.3685,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "codegemma-2b-code-v1-1",
      "display_name": "Codegemma 2B Code V1.1",
      "family": "gemma",
      "org": null,
      "params_b": 2.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "gemma",
      "quantization": "Q4_K_M",
      "ollama_ref": "codegemma:2b-code-v1.1-q4_K_M",
      "size_bytes": 1600000000,
      "model_version_id": 73,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": 15.2962,
      "ranked": true,
      "unranked_reason": null,
      "unranked_reasons": [],
      "fell_for_gotchas": 14,
      "judge_a_mean": 1.3048,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 0.6926,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.7355,
      "agreement": "high",
      "disputed": 13,
      "tokens_per_s": 10.4942,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "code",
      "specialised_label": "code",
      "withheld": null,
      "ci_low": 11.71,
      "ci_high": 19.12,
      "hard_score": 9.1784,
      "hard_n": 71,
      "ext_overall": 15.9223,
      "n_truncated": 92,
      "n_extended": 92,
      "n_gated": 0,
      "beats": [],
      "has_stats": true,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": 227,
      "class_rank": 49,
      "rank_lo": 225,
      "rank_hi": 227,
      "class_rank_lo": 47,
      "class_rank_hi": 49,
      "categories": {
        "reasoning": 17.1429,
        "math": 0.0,
        "coding": 3.9394,
        "instruction": 29.5,
        "knowledge": 16.6667,
        "long_context": 0.0,
        "multilingual": 16.6667,
        "tool_use": 10.0,
        "safety": 43.75,
        "sycophancy": 33.3333,
        "gotchas": 6.6667,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.0394,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.3333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.4,
          "delta_mean": 0.1333,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.0667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.295,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.6333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.3,
          "delta_mean": 1.3333,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1667,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.1714,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.0571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9714,
          "delta_mean": 1.0857,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.4375,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.625,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 1.5417,
          "delta_mean": 0.5,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 1,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.1,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 0.875,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.25,
          "delta_mean": 0.625,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 29.0816,
        "decode_tps_c4": 43.2579,
        "ttft_s": 0.1418,
        "prefill_tps_4k": 858.209,
        "prefill_tps_12k": 856.1354,
        "load_seconds": 10.782,
        "size_gb": 3.3356,
        "vram_gb": 3.3356,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "swiss-ai-apertus-8b-instruct-2509",
      "display_name": "Apertus-8B-Instruct-2509",
      "family": "other",
      "org": "swiss-ai",
      "params_b": 8.05,
      "arch": "dense",
      "hf_downloads": 453286,
      "hf_rank": 337,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "hf.co/unsloth/Apertus-8B-Instruct-2509-GGUF:Q4_K_M",
      "size_bytes": null,
      "model_version_id": 63,
      "run_id": "r20261006-bv1-53e146",
      "battery_version": "v1",
      "tested_at": "2026-10-11T06:48:56.004Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: the model stopped after one token on 212 of 270 prompts in our Ollama import, which points to a broken chat template rather than the model. It was re-run with a chat template written from the official one, on the Vulkan backend; those results replace these.",
      "unranked_reasons": [
        "Withheld: the model stopped after one token on 212 of 270 prompts in our Ollama import, which points to a broken chat template rather than the model. It was re-run with a chat template written from the official one, on the Vulkan backend; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 2.29,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 1.9283,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.535,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 9.4023,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 1,
      "concurrency_max": 1,
      "concurrency_text": "1 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.34.4, flash attention, 16k context; 4 parallel slots until 2026-10-07 00:45Z, 8 after)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: the model stopped after one token on 212 of 270 prompts in our Ollama import, which points to a broken chat template rather than the model. It was re-run with a chat template written from the official one, on the Vulkan backend; those results replace these.",
        "since": "2026-10-12"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 0,
      "n_extended": 0,
      "n_gated": 0,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 2,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.3723,
        "decode_tps_c4": 8.7309,
        "ttft_s": 0.4224,
        "prefill_tps_4k": 293.8039,
        "prefill_tps_12k": 243.2448,
        "load_seconds": 115.3286,
        "size_gb": 22.2718,
        "vram_gb": 22.2718,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon2-11b",
      "display_name": "Falcon2 11B",
      "family": "falcon",
      "org": null,
      "params_b": 11.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon2:11b-q4_K_M",
      "size_bytes": 6800000000,
      "model_version_id": 95,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "long_context partial: 12/18 items scored, 80% must be scored",
      "unranked_reasons": [
        "long_context partial: 12/18 items scored, 80% must be scored"
      ],
      "fell_for_gotchas": 10,
      "judge_a_mean": 3.7281,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.769,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1652,
      "agreement": "high",
      "disputed": 16,
      "tokens_per_s": 3.0053,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 2048,
      "context_tokens": 2048,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": null,
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 9,
      "n_extended": 9,
      "n_gated": 0,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": true,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": 22.8571,
        "math": 16.0,
        "coding": 16.8972,
        "instruction": 57.6667,
        "knowledge": 70.0,
        "long_context": 0.0,
        "multilingual": 22.5,
        "tool_use": 20.0,
        "safety": 87.5,
        "sycophancy": 33.3333,
        "gotchas": 33.3333,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.169,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 1.3667,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 0.9667,
          "delta_mean": 0.4,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.5767,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.8333,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.9,
          "delta_mean": 1.9333,
          "disputed": 10,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": 0.7,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 12,
          "prog_mean": 0.0,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": 0.16,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.225,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": 0.2286,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 2.8571,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.2286,
          "delta_mean": 0.7429,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": 0.875,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 5.5833,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 5.375,
          "delta_mean": 1.125,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": 0.3333,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": 0.2,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": 4.0,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": 2.375,
          "delta_mean": 1.625,
          "disputed": 2,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 8.1448,
        "decode_tps_c4": 19.728,
        "ttft_s": 0.5705,
        "prefill_tps_4k": 184.0331,
        "prefill_tps_12k": 187.1082,
        "load_seconds": 31.782,
        "size_gb": 10.8793,
        "vram_gb": 10.8793,
        "gpu_fraction": 1.0,
        "context_length": 2048.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-10b-instruct",
      "display_name": "Falcon3 10B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 10.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3:10b-instruct-q4_K_M",
      "size_bytes": 6300000000,
      "model_version_id": 96,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (16 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (16 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 5.896,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.3307,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8362,
      "agreement": "high",
      "disputed": 7,
      "tokens_per_s": 2.5913,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (16 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 0,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 9.1919,
        "decode_tps_c4": 21.1544,
        "ttft_s": 0.518,
        "prefill_tps_4k": 176.6338,
        "prefill_tps_12k": 154.0211,
        "load_seconds": 142.203,
        "size_gb": 28.2652,
        "vram_gb": 28.2652,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-1b-instruct",
      "display_name": "Falcon3 1B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 1.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3:1b-instruct-q4_K_M",
      "size_bytes": 1100000000,
      "model_version_id": 97,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (42 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (42 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 3.5034,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.0169,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.8668,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 24.6817,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 8192,
      "context_tokens": 8192,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (42 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 13,
      "n_extended": 13,
      "n_gated": 3,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 4,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 58.9894,
        "decode_tps_c4": 87.3233,
        "ttft_s": 0.0818,
        "prefill_tps_4k": 1301.2234,
        "prefill_tps_12k": 1327.4951,
        "load_seconds": 36.3066,
        "size_gb": 6.0352,
        "vram_gb": 6.0352,
        "gpu_fraction": 1.0,
        "context_length": 8192.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-3b-instruct",
      "display_name": "Falcon3 3B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 3.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3:3b-instruct-q4_K_M",
      "size_bytes": 2000000000,
      "model_version_id": 98,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (169 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (169 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 3.03,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.5733,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.6089,
      "agreement": "high",
      "disputed": 8,
      "tokens_per_s": 5.2881,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "le4",
      "size_class_label": "3–4B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (169 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 2,
      "n_extended": 2,
      "n_gated": 0,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 8,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 25.9809,
        "decode_tps_c4": 57.8924,
        "ttft_s": 0.154,
        "prefill_tps_4k": 685.4602,
        "prefill_tps_12k": 522.43,
        "load_seconds": 93.828,
        "size_gb": 13.9799,
        "vram_gb": 13.9799,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "falcon3-7b-instruct",
      "display_name": "Falcon3 7B Instruct",
      "family": "falcon",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "falcon3:7b-instruct-q4_K_M",
      "size_bytes": 4600000000,
      "model_version_id": 99,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (6 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (6 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 6.3393,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 5.821,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.0974,
      "agreement": "high",
      "disputed": 12,
      "tokens_per_s": 3.235,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (6 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 1,
      "n_extended": 1,
      "n_gated": 1,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 6,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 3,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.6426,
        "decode_tps_c4": 24.7027,
        "ttft_s": 0.3756,
        "prefill_tps_4k": 278.6717,
        "prefill_tps_12k": 240.2589,
        "load_seconds": 112.218,
        "size_gb": 20.1049,
        "vram_gb": 20.1049,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "internlm2-1-8b-chat-v2-5",
      "display_name": "Internlm2 1.8B Chat V2.5",
      "family": "internlm",
      "org": null,
      "params_b": 1.8,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "other",
      "quantization": "Q4_K_M",
      "ollama_ref": "internlm2:1.8b-chat-v2.5-q4_K_M",
      "size_bytes": 1200000000,
      "model_version_id": 129,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for InternLM2 ends the prompt without the newline after <|im_start|>assistant, so on 31 of 270 prompts the model ended its turn after one token and its reply came back empty (a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for InternLM2 ends the prompt without the newline after <|im_start|>assistant, so on 31 of 270 prompts the model ended its turn after one token and its reply came back empty (a serving bug, not the model). It was re-run with the corrected template; those results replace these."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 4.094,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 3.4148,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 0.9188,
      "agreement": "high",
      "disputed": 9,
      "tokens_per_s": 12.9804,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for InternLM2 ends the prompt without the newline after <|im_start|>assistant, so on 31 of 270 prompts the model ended its turn after one token and its reply came back empty (a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 6,
      "n_extended": 6,
      "n_gated": 6,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 5,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 43.041,
        "decode_tps_c4": 73.2188,
        "ttft_s": 0.1045,
        "prefill_tps_4k": 1048.7912,
        "prefill_tps_12k": 822.4664,
        "load_seconds": 81.9965,
        "size_gb": 14.2841,
        "vram_gb": 14.2841,
        "gpu_fraction": 1.0,
        "context_length": 16384.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-4-reasoning",
      "display_name": "Phi-4-reasoning 14B",
      "family": "phi",
      "org": "microsoft",
      "params_b": 14.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi4-reasoning:14b-q4_K_M",
      "size_bytes": 11117500000,
      "model_version_id": 234,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: in our Ollama build, Phi-4-reasoning's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and 259 of 270 answers were cut off before a final answer (a serving problem, not the model). It will be re-run once the serving setup is fixed.",
      "unranked_reasons": [
        "Withheld: in our Ollama build, Phi-4-reasoning's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and 259 of 270 answers were cut off before a final answer (a serving problem, not the model). It will be re-run once the serving setup is fixed."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": null,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": null,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": null,
      "agreement": null,
      "disputed": null,
      "tokens_per_s": null,
      "n_tampering": null,
      "n_flagged": null,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: in our Ollama build, Phi-4-reasoning's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and 259 of 270 answers were cut off before a final answer (a serving problem, not the model). It will be re-run once the serving setup is fixed.",
        "since": "2026-11-10"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": null,
      "n_extended": null,
      "n_gated": null,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": true,
      "is_latest": true,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "identity": {
          "n_items": 3,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "long_context": {
          "n_items": 18,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "math": {
          "n_items": 25,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "safety": {
          "n_items": 24,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "writing": {
          "n_items": 8,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": null,
        "decode_tps_c4": null,
        "ttft_s": null,
        "prefill_tps_4k": null,
        "prefill_tps_12k": null,
        "load_seconds": null,
        "size_gb": null,
        "vram_gb": null,
        "gpu_fraction": null,
        "context_length": null
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "microsoft-phi-4-reasoning-plus",
      "display_name": "Phi-4-reasoning-plus 14B",
      "family": "phi",
      "org": "microsoft",
      "params_b": 14.7,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "mit",
      "quantization": "Q4_K_M",
      "ollama_ref": "phi4-reasoning:14b-plus-q4_K_M",
      "size_bytes": 11117503233,
      "model_version_id": 235,
      "run_id": "r20261105-bv1-2abe78",
      "battery_version": "v1",
      "tested_at": "2026-11-14T12:26:35.367Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: in our Ollama build, Phi-4-reasoning-plus's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and its answers were cut off before a final answer (a serving problem, not the model). The run was stopped after 50 of 270 answers; it will be re-run once the serving setup is fixed.",
      "unranked_reasons": [
        "Withheld: in our Ollama build, Phi-4-reasoning-plus's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and its answers were cut off before a final answer (a serving problem, not the model). The run was stopped after 50 of 270 answers; it will be re-run once the serving setup is fixed."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": null,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": null,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": null,
      "agreement": null,
      "disputed": null,
      "tokens_per_s": null,
      "n_tampering": null,
      "n_flagged": null,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 16384,
      "context_tokens": 16384,
      "backend": "ROCm 7.1 (Ollama 0.40.13, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "9-16",
      "size_class_label": "9–16B",
      "specialised": null,
      "specialised_label": null,
      "withheld": {
        "reason": "Withheld: in our Ollama build, Phi-4-reasoning-plus's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and its answers were cut off before a final answer (a serving problem, not the model). The run was stopped after 50 of 270 answers; it will be re-run once the serving setup is fixed.",
        "since": "2026-11-10"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": null,
      "n_extended": null,
      "n_gated": null,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": null,
      "withheld_only": true,
      "is_latest": true,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "identity": {
          "n_items": 3,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "long_context": {
          "n_items": 18,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "math": {
          "n_items": 25,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "safety": {
          "n_items": 24,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": true,
          "has_prog": false
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": true,
          "judged": false,
          "has_prog": false
        },
        "writing": {
          "n_items": 8,
          "n_scored": null,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": null,
          "agreement": null,
          "n_judge_resolved": null,
          "n_judge_split": null,
          "n_pending_resolution": null,
          "n_unjudgeable": null,
          "coverage_ok": null,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": null,
        "decode_tps_c4": null,
        "ttft_s": null,
        "prefill_tps_4k": null,
        "prefill_tps_12k": null,
        "load_seconds": null,
        "size_gb": null,
        "vram_gb": null,
        "gpu_fraction": null,
        "context_length": null
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-math-1-5b-instruct",
      "display_name": "Qwen2 Math 1.5B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 1.5,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2-math:1.5b-instruct-q4_K_M",
      "size_bytes": 986000000,
      "model_version_id": 163,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little)."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 2.7681,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 2.0233,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.001,
      "agreement": "high",
      "disputed": 5,
      "tokens_per_s": 16.1212,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "tiny",
      "size_class_label": "Tiny (under 3B)",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 24,
      "n_extended": 24,
      "n_gated": 9,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 1,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 44.1416,
        "decode_tps_c4": 61.0823,
        "ttft_s": 0.1041,
        "prefill_tps_4k": 1220.2629,
        "prefill_tps_12k": 1310.6839,
        "load_seconds": 9.1967,
        "size_gb": 2.0633,
        "vram_gb": 2.0633,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    },
    {
      "slug": "qwen2-math-7b-instruct",
      "display_name": "Qwen2 Math 7B Instruct",
      "family": "qwen",
      "org": null,
      "params_b": 7.0,
      "arch": "dense",
      "hf_downloads": null,
      "hf_rank": null,
      "active_params_b": null,
      "vision": false,
      "derived_from": null,
      "license": "apache-2.0",
      "quantization": "Q4_K_M",
      "ollama_ref": "qwen2-math:7b-instruct-q4_K_M",
      "size_bytes": 4700000000,
      "model_version_id": 164,
      "run_id": "r20261016-bv1-473145",
      "battery_version": "v1",
      "tested_at": "2026-10-26T08:44:57.631Z",
      "overall": null,
      "ranked": false,
      "unranked_reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
      "unranked_reasons": [
        "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little)."
      ],
      "fell_for_gotchas": null,
      "judge_a_mean": 5.365,
      "judge_a_model": "openai/gpt-5.6-luna",
      "judge_b_mean": 4.7317,
      "judge_b_model": "anthropic/claude-haiku-5.5",
      "delta_mean": 1.1205,
      "agreement": "high",
      "disputed": 14,
      "tokens_per_s": 3.5514,
      "n_tampering": 0,
      "n_flagged": 0,
      "concurrency_min": 8,
      "concurrency_max": 8,
      "concurrency_text": "8 at a time",
      "context_tokens_min": 4096,
      "context_tokens": 4096,
      "backend": "ROCm 7.1 (Ollama 0.40.4, flash attention, 16k context; 8 parallel slots)",
      "gpu": "AMD Radeon 8060S (Ryzen AI Max+ 395, 128 GB unified, ~78 GB GPU)",
      "backend_short": "ROCm 7.1",
      "size_class": "4-9",
      "size_class_label": "4–9B",
      "specialised": "math",
      "specialised_label": "math",
      "withheld": {
        "reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
        "since": "2026-10-28"
      },
      "ci_low": null,
      "ci_high": null,
      "hard_score": null,
      "hard_n": null,
      "ext_overall": null,
      "n_truncated": 4,
      "n_extended": 4,
      "n_gated": 8,
      "beats": [],
      "has_stats": false,
      "stats_stale": false,
      "hard_basis": "3b_plus",
      "withheld_only": false,
      "is_latest": false,
      "rank": null,
      "class_rank": null,
      "rank_lo": null,
      "rank_hi": null,
      "class_rank_lo": null,
      "class_rank_hi": null,
      "categories": {
        "reasoning": null,
        "math": null,
        "coding": null,
        "instruction": null,
        "knowledge": null,
        "long_context": null,
        "multilingual": null,
        "tool_use": null,
        "safety": null,
        "sycophancy": null,
        "gotchas": null,
        "writing": null,
        "identity": null
      },
      "category_detail": {
        "coding": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 2,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "gotchas": {
          "n_items": 15,
          "n_scored": 15,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "identity": {
          "n_items": 3,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": false,
          "has_prog": false
        },
        "instruction": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 8,
          "agreement": "medium",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "knowledge": {
          "n_items": 30,
          "n_scored": 30,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "long_context": {
          "n_items": 18,
          "n_scored": 18,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "math": {
          "n_items": 25,
          "n_scored": 25,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "multilingual": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "reasoning": {
          "n_items": 35,
          "n_scored": 35,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "safety": {
          "n_items": 24,
          "n_scored": 24,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 4,
          "agreement": "high",
          "n_judge_resolved": 6,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": true,
          "has_prog": true
        },
        "sycophancy": {
          "n_items": 12,
          "n_scored": 12,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": false,
          "judged": false,
          "has_prog": true
        },
        "tool_use": {
          "n_items": 20,
          "n_scored": 20,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "n/a",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": true,
          "overall": true,
          "judged": false,
          "has_prog": true
        },
        "writing": {
          "n_items": 8,
          "n_scored": 0,
          "prog_mean": null,
          "judge_a_model": "openai/gpt-5.6-luna",
          "judge_a_mean": null,
          "judge_b_model": "anthropic/claude-haiku-5.5",
          "judge_b_mean": null,
          "delta_mean": null,
          "disputed": 0,
          "agreement": "high",
          "n_judge_resolved": 0,
          "n_judge_split": 0,
          "n_pending_resolution": 0,
          "n_unjudgeable": 0,
          "coverage_ok": false,
          "overall": false,
          "judged": true,
          "has_prog": false
        }
      },
      "perf": {
        "decode_tps": 11.1851,
        "decode_tps_c4": 17.4345,
        "ttft_s": 0.3723,
        "prefill_tps_4k": 309.4109,
        "prefill_tps_12k": 313.4693,
        "load_seconds": 21.859,
        "size_gb": 6.5495,
        "vram_gb": 6.5495,
        "gpu_fraction": 1.0,
        "context_length": 4096.0
      },
      "perf_units": {
        "context_length": "tokens",
        "decode_tps": "tok/s",
        "decode_tps_c4": "tok/s",
        "gpu_fraction": "ratio",
        "load_seconds": "s",
        "prefill_tps_12k": "tok/s",
        "prefill_tps_4k": "tok/s",
        "size_gb": "GB",
        "ttft_s": "s",
        "vram_gb": "GB"
      }
    }
  ],
  "withheld": [
    {
      "slug": "swiss-ai-apertus-8b-instruct-2509",
      "display_name": "Apertus-8B-Instruct-2509",
      "run_id": "r20261006-bv1-53e146",
      "reason": "Withheld: the model stopped after one token on 212 of 270 prompts in our Ollama import, which points to a broken chat template rather than the model. It was re-run with a chat template written from the official one, on the Vulkan backend; those results replace these.",
      "since": "2026-10-12"
    },
    {
      "slug": "falcon3-10b-instruct",
      "display_name": "Falcon3 10B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (16 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "since": "2026-10-28"
    },
    {
      "slug": "falcon3-1b-instruct",
      "display_name": "Falcon3 1B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (42 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "since": "2026-10-28"
    },
    {
      "slug": "falcon3-3b-instruct",
      "display_name": "Falcon3 3B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (169 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "since": "2026-10-28"
    },
    {
      "slug": "falcon3-7b-instruct",
      "display_name": "Falcon3 7B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Falcon3 leaves out the newline after the last message, so the model sometimes wrote <|assistant|> itself and its reply came back empty (6 of 270 answers here; a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "since": "2026-10-28"
    },
    {
      "slug": "internlm2-1-8b-chat-v2-5",
      "display_name": "Internlm2 1.8B Chat V2.5",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for InternLM2 ends the prompt without the newline after <|im_start|>assistant, so on 31 of 270 prompts the model ended its turn after one token and its reply came back empty (a serving bug, not the model). It was re-run with the corrected template; those results replace these.",
      "since": "2026-10-28"
    },
    {
      "slug": "microsoft-phi-4-reasoning",
      "display_name": "Phi-4-reasoning 14B",
      "run_id": "r20261105-bv1-2abe78",
      "reason": "Withheld: in our Ollama build, Phi-4-reasoning's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and 259 of 270 answers were cut off before a final answer (a serving problem, not the model). It will be re-run once the serving setup is fixed.",
      "since": "2026-11-10"
    },
    {
      "slug": "microsoft-phi-4-reasoning-plus",
      "display_name": "Phi-4-reasoning-plus 14B",
      "run_id": "r20261105-bv1-2abe78",
      "reason": "Withheld: in our Ollama build, Phi-4-reasoning-plus's reasoning came back mixed into the answer instead of as separate thinking, so the thinking budget never applied and its answers were cut off before a final answer (a serving problem, not the model). The run was stopped after 50 of 270 answers; it will be re-run once the serving setup is fixed.",
      "since": "2026-11-10"
    },
    {
      "slug": "qwen2-math-1-5b-instruct",
      "display_name": "Qwen2 Math 1.5B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
      "since": "2026-10-28"
    },
    {
      "slug": "qwen2-math-7b-instruct",
      "display_name": "Qwen2 Math 7B Instruct",
      "run_id": "r20261016-bv1-473145",
      "reason": "Withheld: Ollama's library chat template for Qwen2-Math differs from the official one (no newline between turns or after the final <|im_start|>assistant, and no default system turn). It was re-run with the official template; those results replace these (the scores changed very little).",
      "since": "2026-10-28"
    }
  ],
  "pending": [
    {
      "reason": "a newer Ollama release that can load it (it crashes while loading in the version we use, on both GPU backends)",
      "names": [
        "Rnj 1 8B Instruct"
      ],
      "n": 1
    }
  ],
  "note": null
}
