← Best models

Best model for coding

Models ranked by their coding score, 0 to 100, as on the category page. Only ranked models with enough scored items are included, and models are grouped by total parameters.

Read the memory figure as an estimate, not a guarantee. The memory a model needs is estimated from its file: the weights plus about 20% for the context of a single conversation. It is not a measurement, and a long conversation or several at once needs more. A model with no file size on record is not counted as fitting any amount of memory. Speeds, and the GPU memory shown on the speed page, were measured on this benchmark's server, not on yours. That server kept 1 to 8 answers in flight at a time and loaded a context window of 16,384 tokens for most models; 42 of the models ranked ran at a shorter context. Speeds were measured on one laptop and are comparable only between models on the same backend. See the settings and the memory method.

All sizes

gpt-oss-120b leads here with 99.9 on a scale of 0 to 100; its Overall is 98.6 (95% interval 96.8–100.0).

All sizes: the top 5 of 227.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 gpt-oss-120b 99.930 of 30 items scored 98.696.8–100.0 1–14 11.0Vulkan 78.4 > 35B
2 Gemma 4 12B IT 98.230 of 30 items scored 98.697.0–99.8 1–12 8.1ROCm 7.1 8.9 9–16B
3 Qwen3-Coder-Nextcode model 97.330 of 30 items scored 96.494.2–98.3 2–42 18.1Vulkan 62.1 > 35B
4 Qwen3.8-27B 96.930 of 30 items scored 95.492.9–97.8 2–53 3.5Vulkan 20.2 16–35B
5 gemma-4-31B-it 96.630 of 30 items scored 99.398.3–100.0 1–5 3.8ROCm 7.1 22.4 16–35B

Tiny (under 3B)

Exaone3.5 2.4B Instruct leads its size class with 64.7 on a scale of 0 to 100; its Overall is 76.3 (95% interval 70.9–81.3).

Tiny (under 3B): the top 5 of 49.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 Exaone3.5 2.4B Instruct 64.730 of 30 items scored 76.370.9–81.3 73–141 100.9ROCm 7.1 1.9 Tiny (under 3B)
2 Lfm3 1.7B 61.530 of 30 items scored 70.265.3–75.3 96–162 42.1ROCm 7.1 1.3 Tiny (under 3B)
3 Granite4 1B H 59.930 of 30 items scored 66.861.4–72.1 119–162 24.9ROCm 7.1 1.9 Tiny (under 3B)
4 YI Coder 1.5B Chatcode model 57.730 of 30 items scored 26.522.1–31.5 213–224 49.5ROCm 7.1 1.2 Tiny (under 3B)
5 Opencoder 1.5B Instructcode model 57.330 of 30 items scored 31.927.6–36.4 209–219 17.4ROCm 7.1 1.7 Tiny (under 3B)

3–4B

Granite4 Micro H leads its size class with 68.8 on a scale of 0 to 100; its Overall is 76.6 (95% interval 72.0–81.4).

3–4B: the top 5 of 26.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 Granite4 Micro H 68.830 of 30 items scored 76.672.0–81.4 72–139 13.4ROCm 7.1 2.3 3–4B
2 Spark-X2.6-4B 65.630 of 30 items scored 85.080.9–89.1 43–106 19.7ROCm 7.1 3.0 3–4B
3 Granite4 3B 64.730 of 30 items scored 74.169.0–79.2 85–145 71.3ROCm 7.1 2.5 3–4B
4 granite-4.1-3b 62.630 of 30 items scored 80.775.9–85.3 66–129 20.1ROCm 7.1 2.5 3–4B
5 Granite4.2 3B 61.630 of 30 items scored 78.273.7–83.1 72–140 24.3ROCm 7.1 2.7 3–4B

4–9B

Gemma4 E4B IT leads its size class with 89.9 on a scale of 0 to 100; its Overall is 95.5 (95% interval 93.0–97.7).

4–9B: the top 5 of 69.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 Gemma4 E4B IT 89.930 of 30 items scored 95.593.0–97.7 2–51 13.2ROCm 7.1 7.9 4–9B
2 Granite4.2 8B 85.130 of 30 items scored 81.977.2–86.3 59–128 11.1ROCm 7.1 6.4 4–9B
3 YI Coder 9B Chatcode model 81.330 of 30 items scored 59.053.7–64.5 150–186 10.5ROCm 7.1 6.4 4–9B
4 Granite4 Tiny H 80.830 of 30 items scored 79.174.1–84.2 64–132 24.2ROCm 7.1 5.0 4–9B
5 Gemma4 E2B IT 80.130 of 30 items scored 93.891.0–96.1 4–61 17.2ROCm 7.1 5.5 4–9B

9–16B

Gemma 4 12B IT leads its size class with 98.2 on a scale of 0 to 100; its Overall is 98.6 (95% interval 97.0–99.8).

9–16B: the top 5 of 25.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 Gemma 4 12B IT 98.230 of 30 items scored 98.697.0–99.8 1–12 8.1ROCm 7.1 8.9 9–16B
2 Ministral 3 14B Instruct (2512) 87.530 of 30 items scored 92.790.0–95.2 9–61 5.6ROCm 7.1 10.9 9–16B
3 phi-4 87.030 of 30 items scored 85.981.9–89.8 39–99 5.0ROCm 7.1 10.9 9–16B
4 Qwen2.5-Coder-14B-Instructcode model 82.730 of 30 items scored 87.784.0–91.3 30–86 5.2ROCm 7.1 10.8 9–16B
5 gemma-3-12b-it 81.830 of 30 items scored 92.789.7–95.4 8–66 6.1ROCm 7.1 9.8 9–16B

16–35B

Qwen3.8-27B leads its size class with 96.9 on a scale of 0 to 100; its Overall is 95.4 (95% interval 92.9–97.8).

16–35B: the top 5 of 49.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 Qwen3.8-27B 96.930 of 30 items scored 95.492.9–97.8 2–53 3.5Vulkan 20.2 16–35B
2 gemma-4-31B-it 96.630 of 30 items scored 99.398.3–100.0 1–5 3.8ROCm 7.1 22.4 16–35B
2 gemma-4-26B-A4B-it 96.630 of 30 items scored 97.895.8–99.3 1–22 15.7ROCm 7.1 21.6 16–35B
4 gpt-oss-20b 95.130 of 30 items scored 95.092.2–97.6 2–60 14.9ROCm 7.1 16.6 16–35B
5 Qwen3.6-35B-A3B 94.130 of 30 items scored 94.691.6–97.2 3–63 22.1Vulkan 28.7 16–35B

> 35B

gpt-oss-120b leads its size class with 99.9 on a scale of 0 to 100; its Overall is 98.6 (95% interval 96.8–100.0).

> 35B: the top 5 of 8.
Place Model Coding (0 to 100) Overall Rank range Decode (tok/s) Memory needed (GB, est.) Size class
1 gpt-oss-120b 99.930 of 30 items scored 98.696.8–100.0 1–14 11.0Vulkan 78.4 > 35B
2 Qwen3-Coder-Nextcode model 97.330 of 30 items scored 96.494.2–98.3 2–42 18.1Vulkan 62.1 > 35B
3 Flint 40B A2B 93.430 of 30 items scored 89.085.1–92.6 14–90 25.9ROCm 7.1 27.6 > 35B
4 GLM-4.5-Air 87.530 of 30 items scored 92.889.7–95.6 6–68 6.0Vulkan 87.6 > 35B
5 Sage-A1.5-Q4_K_M 84.630 of 30 items scored 94.591.8–96.7 5–56 22.6ROCm 7.1 26.4 > 35B

A higher place is not always a clear win. When two models have overlapping rank ranges, this battery cannot tell them apart; see how sure a score is. A dash means not measured or not scored, which is different from 0. Speeds were measured on one laptop and are comparable only between models on the same backend.

More: all the best-model pages, the hardware picker, speed and memory for every model.