Calibration
AA Coding Index → SWE-Rebench
SWE-Rebench is estimated from AA Coding Index with a offset logistic curve fitted on 5 models measured on both: y = 0.0000 + (0.5809 − 0.0000) / (1 + exp(−200.00·(x − 0.4296))), R² = 0.78, cross-validated error 8.9 pp. It is used for 59 estimates.
| Estimated model | AA Coding Index | SWE-Rebench | Source |
|---|---|---|---|
| Apodex 1.1 Mini | 60.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| Celeris-1 | 14.4% | 0.0% | estimated ± 8.9 pp, low confidence |
| Claude 3 Opus | 19.5% | 0.0% | estimated ± 8.9 pp, low confidence |
| Claude Fable 5.1 | 81.6% | 58.1% | estimated ± 8.9 pp, low confidence |
| Command A+ | 27.9% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemini 1.5 Pro | 23.6% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemini 3.5 Flash-Lite | 49.3% | 58.1% | estimated ± 8.9 pp, low confidence |
| Gemini 3.6 Flash | 69.2% | 58.1% | estimated ± 8.9 pp, low confidence |
| Gemini 3.7 Flash | 76.1% | 58.1% | estimated ± 8.9 pp, low confidence |
| Gemini 3.8 Flash | 76.3% | 58.1% | estimated ± 8.9 pp, low confidence |
| Gemma 3 27B | 10.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemma 4 12B | 31.0% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemma 4 26B A4B | 39.3% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemma 4 E2B | 7.2% | 0.0% | estimated ± 8.9 pp, low confidence |
| Gemma 4 E4B | 9.4% | 0.0% | estimated ± 8.9 pp, low confidence |
| GLM-5.2 | 68.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| GLM-5.3 | 74.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| GPT-4.1 nano | 11.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| GPT-4 Turbo | 21.5% | 0.0% | estimated ± 8.9 pp, low confidence |
| GPT-4o mini | 11.4% | 0.0% | estimated ± 8.9 pp, low confidence |
| GPT-5.6 Luna | 71.5% | 58.1% | estimated ± 8.9 pp, low confidence |
| GPT-5.6 Sol | 77.4% | 58.1% | estimated ± 8.9 pp, low confidence |
| GPT-5.6 Terra | 76.7% | 58.1% | estimated ± 8.9 pp, low confidence |
| GPT-6 Astra | 76.9% | 58.1% | estimated ± 8.9 pp, low confidence |
| GPT-OSS 120B | 30.4% | 0.0% | estimated ± 8.9 pp, low confidence |
| GPT-OSS 20B | 20.7% | 0.0% | estimated ± 8.9 pp, low confidence |
| Grok 4.3 | 42.3% | 11.2% | estimated ± 8.9 pp, low confidence |
| Grok 4.5 | 72.5% | 58.1% | estimated ± 8.9 pp, low confidence |
| Grok 4.6 | 76.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| Hy3 | 58.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| K-Exaone | 32.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| Kimi K2.7 Code | 60.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| Kimi K3 | 76.2% | 58.1% | estimated ± 8.9 pp, low confidence |
| LFM2.5-2.6B | 7.7% | 0.0% | estimated ± 8.9 pp, low confidence |
| Ling 2.6 Flash | 25.3% | 0.0% | estimated ± 8.9 pp, low confidence |
| Ling 3.0 Flash | 50.6% | 58.1% | estimated ± 8.9 pp, low confidence |
| Ling 3.0 Flash FP8 | 50.6% | 58.1% | estimated ± 8.9 pp, low confidence |
| Llama 4 Maverick | 16.3% | 0.0% | estimated ± 8.9 pp, low confidence |
| Llama 4 Scout | 8.2% | 0.0% | estimated ± 8.9 pp, low confidence |
| MiMo-V2.5-Pro | 60.2% | 58.1% | estimated ± 8.9 pp, low confidence |
| Mistral Large 3 | 20.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| Mistral Small 4 | 26.6% | 0.0% | estimated ± 8.9 pp, low confidence |
| Mistral Small 4 (Reasoning) | 26.6% | 0.0% | estimated ± 8.9 pp, low confidence |
| Muse Spark 1.1 | 71.3% | 58.1% | estimated ± 8.9 pp, low confidence |
| Muse Spark 1.2 | 72.2% | 58.1% | estimated ± 8.9 pp, low confidence |
| Muse Spark 1.3 | 75.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| Nemotron 3 Nano 30B | 14.4% | 0.0% | estimated ± 8.9 pp, low confidence |
| Nemotron 3 Nano Omni 30B A3B | 13.8% | 0.0% | estimated ± 8.9 pp, low confidence |
| Nemotron 3 Super 100B | 37.7% | 0.0% | estimated ± 8.9 pp, low confidence |
| o1 | 39.7% | 0.1% | estimated ± 8.9 pp, low confidence |
| o1-preview | 34.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| Quasar 438B | 61.2% | 58.1% | estimated ± 8.9 pp, low confidence |
| Qwen3.8-27B | 68.1% | 58.1% | estimated ± 8.9 pp, low confidence |
| Qwen3.8-Flash-Next | 73.1% | 58.1% | estimated ± 8.9 pp, low confidence |
| Qwen3.8 Max Preview | 71.8% | 58.1% | estimated ± 8.9 pp, low confidence |
| Step 3.7 Flash | 39.6% | 0.1% | estimated ± 8.9 pp, low confidence |
| Trinity-Large-Preview | 25.8% | 0.0% | estimated ± 8.9 pp, low confidence |
| Trinity-Large-Thinking | 25.8% | 0.0% | estimated ± 8.9 pp, low confidence |
| Ultravox v0.6 Llama 3.3 70B | 11.9% | 0.0% | estimated ± 8.9 pp, low confidence |