ClinicBench: leaderboard
Clinical language-model evaluation across 17 datasets. The score is the mean of all 17 dataset scores; incomplete submissions have no aggregate. Underlying precision is retained, while the source display rounds to one decimal.
Metric: Average Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 | 56.96 | #438 |
| 2 | Llama 3 70B | 52.7 | #631 |
| 3 | Claude 2 | 46.35 | #761 |
| 4 | GPT-3.5 Turbo | 45.06 | #865 |
| 5 | Llama 2 70B | 39.7 | #1201 |
| 6 | vicuna-13B | 36.48 | #1469 |
| 7 | Llama 2 13B | 36.32 | #1436 |
| 8 | Vicuna-7B | 34.61 | #1542 |
| 9 | Llama 2 7B | 33.68 | #1520 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=clinicbench · How It Works · Data refreshed daily, snapshot 2026-10-20.