Calibration
AA-SciCode → DeepSWE
DeepSWE is estimated from AA-SciCode with a Hill curve fitted on 40 models measured on both: y = 0.0000 + (0.8063 − 0.0000)·x^6.00 / (0.42943^6.00 + x^6.00), R² = 0.39, cross-validated error 7.5 pp. It is used for 33 estimates.
| Estimated model | AA-SciCode | DeepSWE | Source |
|---|---|---|---|
| Solar Pro 4 | 48.0% | 53.3% | estimated ± 7.5 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B | 47.8% | 52.8% | estimated ± 7.5 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B NVFP4 | 47.8% | 52.8% | estimated ± 7.5 pp, low confidence |
| Gemini 2.5 Pro | 46.3% | 49.3% | estimated ± 7.5 pp, low confidence |
| Gemma 4 31B | 45.5% | 47.2% | estimated ± 7.5 pp, low confidence |
| Muse Glimmer 30B | 44.9% | 45.7% | estimated ± 7.5 pp, low confidence |
| Qwen3.5 397B | 44.8% | 45.4% | estimated ± 7.5 pp, low confidence |
| Ling 3.0 Flash VL | 44.2% | 43.8% | estimated ± 7.5 pp, low confidence |
| Step 3.7 Flash | 43.9% | 43.0% | estimated ± 7.5 pp, low confidence |
| K-EXAONE 2.0 | 42.0% | 37.6% | estimated ± 7.5 pp, low confidence |
| A.X K2 | 41.0% | 34.8% | estimated ± 7.5 pp, low confidence |
| Trinity-Large-Preview | 40.6% | 33.6% | estimated ± 7.5 pp, low confidence |
| Trinity-Large-Thinking | 40.6% | 33.6% | estimated ± 7.5 pp, low confidence |
| Gemma 4 26B A4B | 40.0% | 31.9% | estimated ± 7.5 pp, low confidence |
| Nemotron 3.5 Bolt 74B-A9B | 39.7% | 31.0% | estimated ± 7.5 pp, low confidence |
| Qwen3.5-122B-A10B | 39.7% | 31.0% | estimated ± 7.5 pp, low confidence |
| DeepSeek V3 0324 | 39.0% | 29.0% | estimated ± 7.5 pp, low confidence |
| GPT-OSS 20B | 38.9% | 28.7% | estimated ± 7.5 pp, low confidence |
| Mistral Small 4 | 38.8% | 28.4% | estimated ± 7.5 pp, low confidence |
| North Mini Code | 38.8% | 28.4% | estimated ± 7.5 pp, low confidence |
| Command A+ | 38.5% | 27.6% | estimated ± 7.5 pp, low confidence |
| Mercury 2.5 | 38.5% | 27.6% | estimated ± 7.5 pp, low confidence |
| Granite 4.2 30B | 37.8% | 25.6% | estimated ± 7.5 pp, low confidence |
| Mistral Large 3 | 36.6% | 22.3% | estimated ± 7.5 pp, low confidence |
| Qwen3.6-35B-A3B | 36.6% | 22.3% | estimated ± 7.5 pp, low confidence |
| DeepSeek V3 | 35.8% | 20.3% | estimated ± 7.5 pp, low confidence |
| GPT-OSS 120B | 34.0% | 15.9% | estimated ± 7.5 pp, low confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 32.1% | 12.0% | estimated ± 7.5 pp, low confidence |
| Llama 4 Maverick | 31.7% | 11.2% | estimated ± 7.5 pp, low confidence |
| Granite 4.2 8B | 31.5% | 10.9% | estimated ± 7.5 pp, low confidence |
| Mistral Small 3.1 | 27.8% | 5.5% | estimated ± 7.5 pp, low confidence |
| MiniCPM5-2B | 26.3% | 4.0% | estimated ± 7.5 pp, low confidence |
| Solar Pro 3 | 25.5% | 3.4% | estimated ± 7.5 pp, low confidence |
Fitted on
| Model | AA-SciCode | DeepSWE |
|---|---|---|
| Claude Fable 5.1 | 63.1% | 67.4% |
| Claude Fable 5.2 | 67.7% | 74.3% |
| Claude Opus 5 | 56.4% | 68.8% |
| Claude Opus 5.5 | 66.9% | 74.2% |
| Claude Sonnet 5.5 | 61.0% | 71.0% |
| DeepSeek V4.1 Flash 1017 | 51.9% | 77.0% |
| DeepSeek V4 Flash 0731 | 50.3% | 54.4% |
| DeepSeek V4 Pro | 51.0% | 62.7% |
| Gemini 3.6 Flash | 53.4% | 49.0% |
| Gemini 3.7 Flash | 57.2% | 65.3% |
| Gemini 3.8 Flash | 56.6% | 73.8% |
| Gemini 4 Argon | 61.8% | 77.9% |
| GLM-5.3 | 59.0% | 66.9% |
| GPT-5.6 Luna | 53.6% | 67.2% |
| GPT-5.6 Sol | 57.1% | 72.7% |
| GPT-5.6 Terra | 55.0% | 69.6% |
| GPT-6.1 Sol | 54.2% | 71.9% |
| GPT-6 Astra | 56.5% | 74.1% |
| GPT-6 Luna | 54.6% | 66.6% |
| GPT-6 Sol | 57.6% | 68.8% |
| Grok 4.5 | 55.0% | 53.0% |
| Grok 4.6 | 56.5% | 65.9% |
| Grok 4.7 | 57.4% | 71.0% |
| Inkling 2 Small | 52.3% | 72.1% |
| Kimi K3.1 Pro | 60.8% | 75.5% |
| Kimi K3 | 59.5% | 67.5% |
| Ling 3.1 Flash | 54.1% | 59.7% |
| MiMo-V2.6-Flash | 51.3% | 67.9% |
| MiMo-V2.6-Pro | 60.9% | 71.9% |
| MiniMax M3.1 Flash | 53.4% | 73.7% |
| Muse Spark 1.1 | 58.8% | 53.3% |
| Muse Spark 1.2 | 57.4% | 59.3% |
| Muse Spark 1.3 | 58.8% | 75.4% |
| Muse Spark Open | 58.4% | 60.0% |
| Muse Wonder | 60.3% | 78.8% |
| Muse Wonder Full | 62.6% | 79.0% |
| Qwen3.8-27B | 46.6% | 42.2% |
| Qwen3.8-Flash-Next | 50.6% | 58.7% |
| Beam 1.0 501B-A23B | 48.5% | 43.3% |
| Step 5 Preview | 58.9% | 67.7% |