The typical standard errors between pairs of models on this dataset as a function of the absolute accuracy.
The standard error of each model pair against their observed accuracy difference. Pairs below a reference line differ by more than the corresponding number of standard errors, i.e. they are statistically distinguishable at that level.
| model | pass1 | pass@count | win_rate | count | SE(A) | SE_x(A) | SE_pred(A) |
|---|---|---|---|---|---|---|---|
| qwen3-32b | 70 | 85.5 | 36.4 | 9 | 0.42 | 0.35 | 0.23 |
| qwen3-14b | 67.3 | 81.6 | 34.4 | 10 | 0.43 | 0.37 | 0.21 |
| llama-3.1-70B-instruct | 63.9 | 63.9 | 32.3 | 12 | 0.44 | 0.44 | 0 |
| qwen3-8b | 62.6 | 78.2 | 31 | 10 | 0.44 | 0.38 | 0.22 |
| qwen2-72b-instruct | 62.2 | 78.1 | 31 | 4 | 0.44 | 0.35 | 0.27 |
| qwen2.5-coder-32b-instruct | 60.5 | 80 | 29.5 | 8 | 0.45 | 0.36 | 0.26 |
| google_gemma_3_12b_it | 58.7 | 77.6 | 28.2 | 11 | 0.45 | 0.38 | 0.23 |
| deepseek_r1_distill_llama_70b | 57.7 | 75.8 | 27.6 | 9 | 0.45 | 0.37 | 0.25 |
| qwen3-4b | 57.6 | 74.4 | 27.8 | 11 | 0.45 | 0.39 | 0.23 |
| mistralai_mixtral_8x22b_instruct_v0.1 | 51.8 | 82.2 | 24 | 9 | 0.46 | 0.34 | 0.31 |
| deepseek_r1_distill_qwen_32b | 51.5 | 73.5 | 23.6 | 8 | 0.46 | 0.36 | 0.28 |
| qwen2-math-72b-instruct | 51.1 | 79.8 | 23.8 | 8 | 0.46 | 0.34 | 0.31 |
| deepseek_r1_distill_qwen_14b | 48.1 | 75.5 | 21.5 | 12 | 0.46 | 0.35 | 0.29 |
| qwen1.5-72b-chat | 47.5 | 47.5 | 21.7 | 1 | 0.46 | NaN | NaN |
| qwen2.5-coder-14b-instruct | 47.4 | 81.9 | 21.3 | 10 | 0.46 | 0.33 | 0.32 |
| qwen1.5-32b-chat | 46.1 | 75 | 20.6 | 8 | 0.45 | 0.34 | 0.31 |
| llama-3.1-8B-instruct | 45 | 45 | 19.9 | 15 | 0.45 | 0.45 | 0 |
| qwen2-7b-instruct | 44.2 | 78.1 | 19.5 | 12 | 0.45 | 0.34 | 0.3 |
| qwen3-1.7b | 42.9 | 68 | 19.2 | 12 | 0.45 | 0.37 | 0.26 |
| mistralai_mixtral_8x7b_instruct_v0.1 | 42.3 | 75.9 | 18.8 | 10 | 0.45 | 0.33 | 0.31 |
| google_gemma_3_4b_it | 41.5 | 66.2 | 17.9 | 13 | 0.45 | 0.38 | 0.25 |
| mistralai_ministral_8b_instruct_2410 | 39.2 | 78.4 | 16.7 | 11 | 0.45 | 0.31 | 0.32 |
| qwen1.5-14b-chat | 37.9 | 69.6 | 16 | 10 | 0.44 | 0.33 | 0.29 |
| qwen2.5-coder-7b-instruct | 37.7 | 78.4 | 16 | 12 | 0.44 | 0.3 | 0.32 |
| mistralai_mathstral_7b_v0.1 | 36.7 | 79.2 | 15.7 | 12 | 0.44 | 0.29 | 0.33 |
| deepseek_r1_distill_llama_8b | 36.1 | 68.5 | 14.7 | 12 | 0.44 | 0.32 | 0.3 |
| deepseek_r1_distill_qwen_7b | 36.1 | 68.1 | 15.1 | 12 | 0.44 | 0.33 | 0.29 |
| llama-3.2-3B-instruct | 35 | 35 | 15 | 19 | 0.43 | 0.43 | 0 |
| mistralai_mistral_7b_instruct_v0.3 | 33.7 | 71.8 | 14.3 | 12 | 0.43 | 0.31 | 0.3 |
| qwen2-math-7b-instruct | 33.1 | 72.4 | 14.6 | 12 | 0.43 | 0.3 | 0.31 |
| qwen2.5-coder-3b-instruct | 29.4 | 74.6 | 12.1 | 12 | 0.42 | 0.27 | 0.31 |
| mistralai_mistral_7b_instruct_v0.2 | 29.1 | 65.4 | 11.9 | 12 | 0.41 | 0.3 | 0.28 |
| deepseek_v2_lite_chat | 29 | 67.5 | 11.9 | 10 | 0.41 | 0.28 | 0.3 |
| qwen1.5-7b-chat | 25.1 | 62.5 | 10.1 | 10 | 0.4 | 0.26 | 0.3 |
| qwen2-math-1.5b-instruct | 25.1 | 69.3 | 11.5 | 12 | 0.4 | 0.25 | 0.31 |
| qwen3-0.6b | 23.8 | 54 | 10.7 | 13 | 0.39 | 0.29 | 0.26 |
| mistralai_mistral_7b_instruct_v0.1 | 23.8 | 67.2 | 9.96 | 12 | 0.39 | 0.25 | 0.3 |
| llama-3.2-1B-instruct | 21.5 | 21.5 | 9.44 | 21 | 0.37 | 0.37 | 0 |
| deepseek_r1_distill_qwen_1.5b | 20.5 | 59.4 | 8.21 | 12 | 0.37 | 0.22 | 0.29 |
| qwen2.5-coder-1.5b-instruct | 20.3 | 64.8 | 8.48 | 12 | 0.37 | 0.22 | 0.29 |
| qwen2-1.5b-instruct | 17.2 | 63 | 7.59 | 12 | 0.34 | 0.18 | 0.29 |
| qwen1.5-1.8b-chat | 12.4 | 45.6 | 5.71 | 10 | 0.3 | 0.16 | 0.25 |
| qwen2-0.5b-instruct | 11.7 | 55.2 | 6.16 | 13 | 0.29 | 0.13 | 0.26 |
| qwen2.5-coder-0.5b-instruct | 10.4 | 53.1 | 5.79 | 13 | 0.28 | 0.1 | 0.26 |
| qwen1.5-0.5b-chat | 10.3 | 57 | 5.83 | 13 | 0.28 | 0.098 | 0.26 |