healthcare benchmark
HealthBench
An open-source benchmark for measuring performance and safety of large language models in healthcare, consisting of 5,000 multi-turn conversations evaluated by 262 physicians using 48,562 unique rubric criteria across health contexts and behavioral dimensions
Updated Aug 7, 2026
Published models9
Registry coverage9
MetricScore
EvidenceB
HealthBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 60.2% | 100.0% | 9 | C | |
| 2 | MA | 58.0% | 87.5% | 9 | C | |
| 3 | OP | 57.6% | 75.0% | 9 | C | |
| 4 | OP | 57.0% | 62.5% | 9 | C | |
| 5 | OP | 57.0% | 50.0% | 9 | C | |
| 6 | OP | 55.8% | 37.5% | 9 | C | |
| 7 | OP | 54.1% | 25.0% | 9 | C | |
| 8 | OP | 51.4% | 12.5% | 9 | C | |
| 9 | OP | 42.5% | 0.0% | 9 | C |
Score distribution
Top published rows on the benchmark's original scale.
HealthBench highlights
The top published results on this benchmark's own scale.
What is HealthBench?
Definition and scoring fields from the benchmark registry.
An open-source benchmark for measuring performance and safety of large language models in healthcare, consisting of 5,000 multi-turn conversations evaluated by 262 physicians using 48,562 unique rubric criteria across health contexts and behavioral dimensions
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- HealthBench
- Modality
- text
- Primary category
- healthcare
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- healthbench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about HealthBench.
Which model scores highest on HealthBench?
Qwen3.8 Max is currently ranked first with 60.2%.
What does HealthBench measure?
An open-source benchmark for measuring performance and safety of large language models in healthcare, consisting of 5,000 multi-turn conversations evaluated by 262 physicians using 48,562 unique rubric criteria across health contexts and behavioral dimensions
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
9 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.