Calibration
FrontierCode 1.1 Main → Vals LiveCodeBench
Vals LiveCodeBench is estimated from FrontierCode 1.1 Main with a linear curve fitted on 9 models measured on both: y = 0.2703·x + 0.7456, R² = 0.73, cross-validated error 1.9 pp. It is used for 7 estimates.
| Estimated model | FrontierCode 1.1 Main | Vals LiveCodeBench | Source |
|---|---|---|---|
| Claude Haiku 5.5 | 46.4% | 87.1% | estimated ± 1.9 pp, high confidence |
| Claude Opus 4.6 | 26.9% | 81.8% | estimated ± 1.9 pp, high confidence |
| Claude Opus 5.5 | 54.4% | 89.3% | estimated ± 1.9 pp, medium confidence |
| Claude Sonnet 5.5 | 46.2% | 87.0% | estimated ± 1.9 pp, high confidence |
| GPT-6 Astra | 53.3% | 89.0% | estimated ± 1.9 pp, high confidence |
| SWE-1.7 | 42.3% | 86.0% | estimated ± 1.9 pp, high confidence |
| SWE-2 | 50.0% | 88.1% | estimated ± 1.9 pp, high confidence |