HealthBench Hard: leaderboard
Hard subset of HealthBench, evaluating difficult clinical and biomedical advice with physician-written rubrics and stricter scoring.
Metric: Score (%). Source: openai.com. Source: llm-stats.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Spark | 42.8 | #64 |
| 2 | GPT-6.1 Sol | 36.2 | #7 |
| 3 | GPT-5.6 Sol | 33.1 | #19 |
| 4 | GPT-5.6 Terra | 32.7 | #41 |
| 5 | GPT-5.6 Luna | 32 | #68 |
| 6 | GPT-OSS-120B | 30 | #341 |
| 7 | GPT-5.3 | 25.9 | #103 |
| 8 | GPT-5.5 Instant | 22.9 | #129 |
| 9 | Seed 2.0 Mini | 15.3 | #276 |
| 10 | GPT-OSS-20B | 10.8 | #518 |
| 11 | GPT-5 | 1.6 | #98 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=healthbench-hard · How It Works · Data refreshed daily, snapshot 2026-10-20.