Calibration
SuperGPQA → Artificial Analysis Intelligence Index
Artificial Analysis Intelligence Index is estimated from SuperGPQA with a offset logistic curve fitted on 12 models measured on both: y = 0.1193 + (0.2666 − 0.1193) / (1 + exp(−47.14·(x − 0.6335))), R² = 0.81, cross-validated error 3.0 pp. It is used for 10 estimates.
| Estimated model | SuperGPQA | Artificial Analysis Intelligence Index | Source |
|---|---|---|---|
| Qwen3.7 Plus | 71.4% | 26.3% | estimated ± 3.0 pp, high confidence |
| Seed 2.1 Pro | 70.8% | 26.2% | estimated ± 3.0 pp, high confidence |
| Qwen3.5 397B | 70.4% | 26.1% | estimated ± 3.0 pp, high confidence |
| Seed 2.1 Turbo | 67.4% | 24.8% | estimated ± 3.0 pp, high confidence |
| Qwen3.5-122B-A10B | 67.1% | 24.5% | estimated ± 3.0 pp, high confidence |
| Qwen3 235B 2507 (Reasoning) | 64.9% | 21.9% | estimated ± 3.0 pp, high confidence |
| Qwen3.6-35B-A3B | 64.7% | 21.6% | estimated ± 3.0 pp, high confidence |
| Qwen3 235B 2507 | 62.6% | 18.0% | estimated ± 3.0 pp, high confidence |
| MiniCPM5-2B | 40.8% | 11.9% | estimated ± 3.0 pp, medium confidence |
| MiniCPM5-1B | 23.1% | 11.9% | estimated ± 3.0 pp, medium confidence |
Fitted on
| Model | SuperGPQA | Artificial Analysis Intelligence Index |
|---|---|---|
| Claude Opus 4.5 | 70.6% | 23.7% |
| Claude Opus 4.6 | 95.0% | 26.4% |
| Claude Sonnet 4.6 | 95.0% | 24.7% |
| GLM-5 | 66.8% | 27.9% |
| Kimi K2 | 57.2% | 12.7% |
| Kimi K2.5 | 69.2% | 23.5% |
| Qwen3.5-27B | 65.6% | 22.9% |
| Qwen3.5-35B-A3B | 63.4% | 19.3% |
| Qwen3.6-27B | 66.0% | 21.4% |
| Qwen 3.6 Max (preview) | 73.9% | 28.4% |
| Qwen3.6 Plus | 71.6% | 27.0% |
| Qwen3.7 Max | 73.6% | 29.5% |