Medmarks - HealthBench Hard — leaderboard
Metric: Score (%). Source: medmarks.ai. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (Medium) | 44.95 |
| 2 | GPT-5.1 (Medium) | 43.32 |
| 3 | GPT-OSS-120B (High) | 36.2 |
| 4 | Qwen 3 235B A22B (Thinking) | 34.48 |
| 5 | Qwen 3 30B A3B (Thinking) | 28.07 |
| 6 | Gemini 3 Pro (Preview) | 27.74 |
| 7 | GPT-OSS-20B (High) | 26.15 |
| 8 | Claude Sonnet 4.5 | 25.85 |
| 9 | GLM-4.7 FP8 | 25.13 |
| 10 | Qwen 3 8B (Thinking) | 24.67 |
Interactive version: theaggregate.ai/benchmark?slug=medmarks-healthbench-hard · How the rankings work · Data refreshed daily, snapshot 2026-07-22.