benchgap
Calibration

HealthBench Hard → GPQA

GPQA is estimated from HealthBench Hard with a linear curve fitted on 6 models measured on both: y = 0.1148·x + 0.8970, R² = 0.35, cross-validated error 1.7 pp. It is used for 0 estimates.

Fitted on

ModelHealthBench HardGPQA
Claude Opus 4.614.8%91.3%
GPT-5.440.1%92.8%
GPT-5.6 Luna32.0%92.3%
GPT-5.6 Sol33.1%94.6%
GPT-5.6 Terra32.7%92.9%
GPT-6 Astra36.6%96.0%