Calibration
GPQA → Artificial Analysis Intelligence Index
Artificial Analysis Intelligence Index is estimated from GPQA with a offset logistic curve fitted on 62 models measured on both: y = 0.1027 + (1.1050 − 0.1027) / (1 + exp(−17.79·(x − 0.9729))), R² = 0.79, cross-validated error 6.2 pp. It is used for 10 estimates.
| Estimated model | GPQA | Artificial Analysis Intelligence Index | Source |
|---|---|---|---|
| Claude 3.5 Sonnet | 59.4% | 10.4% | estimated ± 6.2 pp, medium confidence |
| Claude Sonnet 4.5 | 83.4% | 18.1% | estimated ± 6.2 pp, medium confidence |
| Interfaze Beta | 89.9% | 31.5% | estimated ± 6.2 pp, medium confidence |
| LFM2.5-230M | 25.4% | 10.3% | estimated ± 6.2 pp, low confidence |
| LFM2.5-VL-450M | 25.7% | 10.3% | estimated ± 6.2 pp, low confidence |
| MAI-Thinking-1 | 84.2% | 19.2% | estimated ± 6.2 pp, medium confidence |
| Qwen3.8-Omni-Flash | 91.0% | 34.9% | estimated ± 6.2 pp, medium confidence |
| Soofi S 30B-A3B | 43.4% | 10.3% | estimated ± 6.2 pp, medium confidence |
| ZAYA1-74B-Preview | 57.3% | 10.3% | estimated ± 6.2 pp, medium confidence |
| ZAYA1-8B | 71.0% | 11.2% | estimated ± 6.2 pp, medium confidence |