# HealthBench Hard → GPQA

Calibration · https://benchgap.net/calibration/healthbench-hard/current/to/gpqa/current · data retrieved 2026-11-21

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

| Model | HealthBench Hard | GPQA |
|---|---|---|
| [Claude Opus 4.6](https://benchgap.net/model/claude-opus-4-6.md) | 14.8% | 91.3% |
| [GPT-5.4](https://benchgap.net/model/gpt-5-4.md) | 40.1% | 92.8% |
| [GPT-5.6 Luna](https://benchgap.net/model/gpt-5-6-luna.md) | 32.0% | 92.3% |
| [GPT-5.6 Sol](https://benchgap.net/model/gpt-5-6-sol.md) | 33.1% | 94.6% |
| [GPT-5.6 Terra](https://benchgap.net/model/gpt-5-6-terra.md) | 32.7% | 92.9% |
| [GPT-6 Astra](https://benchgap.net/model/gpt-6-astra.md) | 36.6% | 96.0% |
