Calibration
FrontierMath v2 (Tier 4) → FrontierMath v2 (Tiers 1-3)
FrontierMath v2 (Tiers 1-3) is estimated from FrontierMath v2 (Tier 4) with a Michaelis–Menten + offset curve fitted on 44 models measured on both: y = 0.1164 + 1.5168·x / (0.84044 + x), R² = 0.89, cross-validated error 7.4 pp. It is used for 1 estimate.
| Estimated model | FrontierMath v2 (Tier 4) | FrontierMath v2 (Tiers 1-3) | Source |
|---|---|---|---|
| GPT-6 Astra | 97.6% | 93.1% | estimated ± 7.4 pp, low confidence |