Calibration
GPQA → ARC-AGI-3
ARC-AGI-3 is estimated from GPQA with a offset logistic curve fitted on 12 models measured on both: y = 0.0000 + (0.7656 − 0.0000) / (1 + exp(−200.00·(x − 0.9548))), R² = 0.99, cross-validated error 9.3 pp. It is used for 4 estimates.
| Estimated model | GPQA | ARC-AGI-3 | Source |
|---|---|---|---|
| Qwen3.5 397B | 88.4% | 0.0% | estimated ± 9.3 pp, low confidence |
| Qwen3.6-27B | 87.8% | 0.0% | estimated ± 9.3 pp, low confidence |
| Qwen3.5-122B-A10B | 86.6% | 0.0% | estimated ± 9.3 pp, low confidence |
| Qwen3.6-35B-A3B | 86.0% | 0.0% | estimated ± 9.3 pp, low confidence |
Fitted on
| Model | GPQA | ARC-AGI-3 |
|---|---|---|
| Claude Opus 4.6 | 91.3% | 0.5% |
| Claude Opus 4.7 | 94.2% | 0.2% |
| Claude Opus 4.8 | 93.6% | 1.5% |
| GPT-5.4 | 92.8% | 0.2% |
| GPT-5.5 | 93.6% | 0.4% |
| GPT-5.6 Luna | 92.3% | 0.2% |
| GPT-5.6 Sol | 94.6% | 7.8% |
| GPT-5.6 Terra | 92.9% | 0.8% |
| GPT-6.2 Astra | 97.0% | 70.3% |
| GPT-6.2 Sol | 96.0% | 54.9% |
| GPT-6 Astra | 96.0% | 62.7% |
| Inkling 2 Small | 93.0% | 0.3% |