healthcare benchmark
HealthBench Hard
A challenging variation of HealthBench that evaluates large language models' performance and safety in healthcare through 5,000 multi-turn conversations with particularly rigorous evaluation criteria validated by 262 physicians from 60 countries
Updated Aug 7, 2026
Published models9
Registry coverage9
MetricScore
EvidenceC
HealthBench Hard leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | ME | 42.8% | 100.0% | 9 | C | |
| 2 | OP | 33.1% | 87.5% | 9 | C | |
| 3 | OP | 32.7% | 75.0% | 9 | C | |
| 4 | OP | 32.0% | 62.5% | 9 | C | |
| 5 | OP | 30.0% | 50.0% | 9 | C | |
| 6 | OP | 25.9% | 37.5% | 9 | C | |
| 7 | OP | 22.9% | 25.0% | 9 | C | |
| 8 | OP | 10.8% | 12.5% | 9 | C | |
| 9 | OP | 1.6% | 0.0% | 9 | C |
Score distribution
Top published rows on the benchmark's original scale.
HealthBench Hard highlights
The top published results on this benchmark's own scale.
What is HealthBench Hard?
Definition and scoring fields from the benchmark registry.
A challenging variation of HealthBench that evaluates large language models' performance and safety in healthcare through 5,000 multi-turn conversations with particularly rigorous evaluation criteria validated by 262 physicians from 60 countries
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level C.
- Family
- HealthBench Hard
- Modality
- text
- Primary category
- healthcare
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- healthbench-hard|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about HealthBench Hard.
Which model scores highest on HealthBench Hard?
Muse Spark is currently ranked first with 42.8%.
What does HealthBench Hard measure?
A challenging variation of HealthBench that evaluates large language models' performance and safety in healthcare through 5,000 multi-turn conversations with particularly rigorous evaluation criteria validated by 262 physicians from 60 countries
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.