Calibration
AA Coding Index → SWE-bench Verified
SWE-bench Verified is estimated from AA Coding Index with a Hill curve fitted on 32 models measured on both: y = 0.0000 + (0.9427 − 0.0000)·x^2.17 / (0.25575^2.17 + x^2.17), R² = 0.91, cross-validated error 5.5 pp. It is used for 35 estimates.
| Estimated model | AA Coding Index | SWE-bench Verified | Source |
|---|---|---|---|
| Apodex 1.1 Mini | 60.8% | 81.8% | estimated ± 5.5 pp, medium confidence |
| Celeris-1 | 14.4% | 21.0% | estimated ± 5.5 pp, low confidence |
| Claude 3 Opus | 19.5% | 33.7% | estimated ± 5.5 pp, low confidence |
| Command A+ | 27.9% | 51.5% | estimated ± 5.5 pp, medium confidence |
| Gemini 1.5 Pro | 23.6% | 43.1% | estimated ± 5.5 pp, medium confidence |
| Gemma 3 27B | 10.1% | 11.0% | estimated ± 5.5 pp, low confidence |
| Gemma 4 12B | 31.0% | 56.8% | estimated ± 5.5 pp, medium confidence |
| Gemma 4 26B A4B | 39.3% | 67.7% | estimated ± 5.5 pp, medium confidence |
| Gemma 4 E2B | 7.2% | 5.7% | estimated ± 5.5 pp, low confidence |
| Gemma 4 E4B | 9.4% | 9.6% | estimated ± 5.5 pp, low confidence |
| GPT-4.1 nano | 11.1% | 13.3% | estimated ± 5.5 pp, low confidence |
| GPT-4 Turbo | 21.5% | 38.3% | estimated ± 5.5 pp, medium confidence |
| GPT-4o mini | 11.4% | 13.9% | estimated ± 5.5 pp, low confidence |
| GPT-OSS 120B | 30.4% | 55.9% | estimated ± 5.5 pp, medium confidence |
| GPT-OSS 20B | 20.7% | 36.5% | estimated ± 5.5 pp, medium confidence |
| Hy3 | 58.8% | 81.0% | estimated ± 5.5 pp, medium confidence |
| K-Exaone | 32.1% | 58.5% | estimated ± 5.5 pp, medium confidence |
| LFM2.5-2.6B | 7.7% | 6.6% | estimated ± 5.5 pp, low confidence |
| Ling 2.6 Flash | 25.3% | 46.5% | estimated ± 5.5 pp, medium confidence |
| Ling 3.0 Flash FP8 | 50.6% | 76.8% | estimated ± 5.5 pp, medium confidence |
| Llama 4 Maverick | 16.3% | 25.7% | estimated ± 5.5 pp, low confidence |
| Llama 4 Scout | 8.2% | 7.3% | estimated ± 5.5 pp, low confidence |
| Mistral Large 3 | 20.1% | 35.0% | estimated ± 5.5 pp, low confidence |
| Mistral Small 4 | 26.6% | 49.2% | estimated ± 5.5 pp, medium confidence |
| Mistral Small 4 (Reasoning) | 26.6% | 49.2% | estimated ± 5.5 pp, medium confidence |
| Nemotron 3 Nano 30B | 14.4% | 21.0% | estimated ± 5.5 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 13.8% | 19.5% | estimated ± 5.5 pp, low confidence |
| Nemotron 3 Super 100B | 37.7% | 65.9% | estimated ± 5.5 pp, medium confidence |
| o1 | 39.7% | 68.1% | estimated ± 5.5 pp, medium confidence |
| o1-preview | 34.1% | 61.3% | estimated ± 5.5 pp, medium confidence |
| Quasar 438B | 61.2% | 81.9% | estimated ± 5.5 pp, medium confidence |
| Qwen3.8 Max Preview | 71.8% | 85.2% | estimated ± 5.5 pp, medium confidence |
| Trinity-Large-Preview | 25.8% | 47.5% | estimated ± 5.5 pp, medium confidence |
| Trinity-Large-Thinking | 25.8% | 47.5% | estimated ± 5.5 pp, medium confidence |
| Ultravox v0.6 Llama 3.3 70B | 11.9% | 15.1% | estimated ± 5.5 pp, low confidence |