Calibration
AA Coding Index → AA-SciCode
AA-SciCode is estimated from AA Coding Index with a linear curve fitted on 78 models measured on both: y = 0.4473·x + 0.2248, R² = 0.87, cross-validated error 3.2 pp. It is used for 21 estimates.
| Estimated model | AA Coding Index | AA-SciCode | Source |
|---|---|---|---|
| Claude 3 Opus | 19.5% | 31.2% | estimated ± 3.2 pp, high confidence |
| Claude Opus 4.7 (Adaptive) | 73.6% | 55.4% | estimated ± 3.2 pp, high confidence |
| Gemini 1.5 Pro | 23.6% | 33.1% | estimated ± 3.2 pp, high confidence |
| Gemma 4 12B | 31.0% | 36.3% | estimated ± 3.2 pp, high confidence |
| Gemma 4 E2B | 7.2% | 25.7% | estimated ± 3.2 pp, medium confidence |
| GLM-4.7 | 45.3% | 42.7% | estimated ± 3.2 pp, high confidence |
| GPT-4.1 mini | 20.2% | 31.5% | estimated ± 3.2 pp, high confidence |
| GPT-4.1 nano | 11.1% | 27.5% | estimated ± 3.2 pp, high confidence |
| GPT-4 Turbo | 21.5% | 32.1% | estimated ± 3.2 pp, high confidence |
| GPT-4o mini | 11.4% | 27.6% | estimated ± 3.2 pp, high confidence |
| GPT-5.1 | 49.4% | 44.6% | estimated ± 3.2 pp, high confidence |
| GPT-5.4 | 71.1% | 54.3% | estimated ± 3.2 pp, high confidence |
| GPT-5 (high) | 37.8% | 39.4% | estimated ± 3.2 pp, high confidence |
| K-Exaone | 32.1% | 36.8% | estimated ± 3.2 pp, high confidence |
| Kimi K2.5 (Reasoning) | 46.8% | 43.4% | estimated ± 3.2 pp, high confidence |
| MiMo-V2-Flash | 49.8% | 44.8% | estimated ± 3.2 pp, high confidence |
| Muse Spark | 58.6% | 48.7% | estimated ± 3.2 pp, high confidence |
| o1 | 39.7% | 40.2% | estimated ± 3.2 pp, high confidence |
| o1-preview | 34.1% | 37.7% | estimated ± 3.2 pp, high confidence |
| Qwen3.6 Plus | 54.5% | 46.9% | estimated ± 3.2 pp, high confidence |
| Ultravox v0.6 Llama 3.3 70B | 11.9% | 27.8% | estimated ± 3.2 pp, high confidence |