HealthBench Hard
HealthBench Hard - a challenging subset of HealthBench conversations testing realistic health conversations with individuals and health professionals.
Category
General Capabilities
Max Score
100.0
Score Type
percent
Active
Yes
Model Scores
Scores for HealthBench Hard
| Model | Score | Date | Verified |
|---|---|---|---|
| GPT-OSS-120B |
100.00%
|
08.08.2025 | Verified |
| GPT-OSS-20B |
10.80
|
05.08.2025 | Verified |