Calibration
SuperGPQA → AA-Omniscience Accuracy
AA-Omniscience Accuracy is estimated from SuperGPQA with a Michaelis–Menten curve fitted on 12 models measured on both: y = 2.0000·x / (3.92058 + x), R² = 0.46, cross-validated error 6.7 pp. It is used for 8 estimates.
| Estimated model | SuperGPQA | AA-Omniscience Accuracy | Source |
|---|---|---|---|
| Seed 2.1 Pro | 70.8% | 30.6% | estimated ± 6.7 pp, low confidence |
| Seed 2.1 Turbo | 67.4% | 29.3% | estimated ± 6.7 pp, low confidence |
| Qwen3.5-122B-A10B | 67.1% | 29.2% | estimated ± 6.7 pp, low confidence |
| Qwen3 235B 2507 (Reasoning) | 64.9% | 28.4% | estimated ± 6.7 pp, low confidence |
| Qwen3.6-35B-A3B | 64.7% | 28.3% | estimated ± 6.7 pp, low confidence |
| Qwen3 235B 2507 | 62.6% | 27.5% | estimated ± 6.7 pp, low confidence |
| MiniCPM5-2B | 40.8% | 18.9% | estimated ± 6.7 pp, low confidence |
| MiniCPM5-1B | 23.1% | 11.1% | estimated ± 6.7 pp, low confidence |
Fitted on
| Model | SuperGPQA | AA-Omniscience Accuracy |
|---|---|---|
| Claude Opus 4.5 | 70.6% | 40.9% |
| Claude Opus 4.6 | 95.0% | 45.8% |
| Claude Sonnet 4.6 | 95.0% | 38.6% |
| GLM-5 | 66.8% | 26.3% |
| Kimi K2 | 57.2% | 27.4% |
| Kimi K2.5 | 69.2% | 35.2% |
| Qwen3.5-27B | 65.6% | 20.7% |
| Qwen3.5-35B-A3B | 63.4% | 20.1% |
| Qwen3.6-27B | 66.0% | 19.6% |
| Qwen 3.6 Max (preview) | 73.9% | 37.9% |
| Qwen3.6 Plus | 71.6% | 26.4% |
| Qwen3.7 Max | 73.6% | 31.1% |