LLMEval-Med - Medical Reasoning: leaderboard
LLMEval-Med test questions on clinical inference: symptom diagnosis, research interpretation, efficacy evaluation and treatment planning, mostly in Chinese and open-ended. A response is usable when a GPT-4o judge with an expert-written checklist averages 4 or more of 5 over three judgements; usability rate in percent, higher is better. Zhang et al. (Fudan University), Findings of EMNLP 2025.
Metric: Usability rate: share of responses a GPT-4o judge scores 4 or 5 of 5 against an expert checklist (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 63.4 | #467 |
| 2 | O1 Preview | 62.75 | #483 |
| 3 | O1 Mini | 60.4 | #761 |
| 4 | GPT-4o | 55.23 | #741 |
| 5 | DeepSeek V3 | 38.24 | #705 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=llmeval-med-medical-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.