Med-R2: leaderboard
Med-R2 (2026) tests whether medical vision-language models reason from visual evidence along the clinical workflow (image quality, anatomy, lesion, then a joint question chain) and resist misleading positive or negative cues injected into the questions. The score is the mean accuracy (%) over its four tasks; higher is better.
Metric: Overall average accuracy (%) over the Image Quality, Anatomy, Lesion and Joint QA tasks of Med-R2, zero-shot with system prompts; each task mixes equal numbers of positive, negative and neutral adversarial question forms; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.2 | Thinking | 67.4 | #185 |
| 2 | InternVL3.5-8B | 61.8 | #1604 | |
| 3 | Hulu-Med-14B | 61.8 | #1607 | |
| 4 | Qwen 3 VL 8B | 61 | #917 | |
| 5 | Qwen 3 VL 235B A22B | 60.6 | #485 | |
| 6 | Lingshu-7B | 56.3 | #1635 | |
| 7 | HuaTuoGPT-Vision-34B | 55.6 | #1828 | |
| 8 | GPT-4o | 55.4 | #751 | |
| 9 | Hulu-Med-4B | 54.5 | #2392 | |
| 10 | Qwen 2.5 VL 7B Instruct | 53.3 | #1902 | |
| 11 | Janus-Pro-7B | 52 | #6857 | |
| 12 | HuluMed-7B | 51.2 | #2049 | |
| 13 | HuatuoGPT-Vision-7B | 50.5 | #2162 | |
| 14 | MedGemma-4B | 34.6 | #2557 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=med-r2 · How It Works · Data refreshed daily, snapshot 2026-11-21.