HealthBench: leaderboard
HealthBench: Evaluates clinical, biomedical, medical-exam, coding, or healthcare-document reasoning.
Metric: Score (%). Source: openai.com. Status: saturated. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 | 60 |
| 2 | GPT-4.1 | 48 |
| 3 | Grok 3 | 46 |
| 4 | Gemini 2.5 Pro | 44 |
| 5 | O1 | 42 |
| 6 | GPT-4.1 Nano | 34 |
| 7 | GPT-4o | 32 |
| 8 | Claude 3.7 Sonnet | 30 |
| 9 | Llama 4 Maverick | 26 |
| 10 | GPT-3.5 Turbo | 16 |
Interactive version: theaggregate.ai/benchmark?slug=healthbench · How It Works · Data refreshed daily, snapshot 2026-09-05.