HealthBench Hard (updated implementation)
Professional Work · 2026-04-23
HealthBench Hard re-scored under the updated HealthBench implementation OpenAI introduced with the GPT-5.5 system card in April 2026, under which it recomputed prior models. Not comparable with Hard scores OpenAI published before 2026-04-23.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5 | 41.6 |
| GPT-5.1 | 41.4 |
| GPT-5.2 | 38.9 |
| GPT-6 Luna (ChatGPT, October 2026) | 36.1 |
| GPT-5.6 Terra | 34.3 |
| GPT-6 Astra | 34.2 |
| GPT-5.5 | 33.8 |
| GPT-6.1 Sol | 33.4 |
| GPT-5.6 Luna | 31.4 |
| GPT-5.6 Sol | 31.1 |
| GPT-6 Sol (ChatGPT, October 2026) | 30.9 |
| GPT-5.4 | 30.3 |
| GPT-5.6 Sol (ChatGPT, August 2026) | 27.1 |
| GPT-6 Luna | 25.4 |
| GPT-5.6 Luna (ChatGPT, August 2026) | 24.9 |
| GPT-5.2 Instant (source unspecified) | 23.5 |
| GPT-5.1 Instant | 23.0 |
| GPT-6 Sol | 22.1 |
| GPT-5.5 Instant | 21.3 |
| GPT-5.3 Instant (source unspecified) | 17.8 |