Med-R2 - Image Quality: leaderboard

Accuracy (%) of medical vision-language models on the Med-R2 (2026) image quality task, judging the quality and acquisition of the medical image, with equal numbers of positive, negative and neutral adversarial question forms; higher is better.

Metric: Accuracy (%) on the Med-R2 Image Quality task (judging the quality and acquisition of the medical image), 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

#ModelSettingScoreOverall rank
1Janus-Pro-7B74.6#6857
2Lingshu-7B67.3#1635
3GPT-5.2Thinking63.7#185
4InternVL3.5-8B58.3#1604
5Qwen 3 VL 235B A22B54.2#485
6Qwen 3 VL 8B53.3#917
7GPT-4o53#751
8Qwen 2.5 VL 7B Instruct49.3#1902
9HuaTuoGPT-Vision-34B48#1828
10Hulu-Med-14B47.4#1607
11HuatuoGPT-Vision-7B41.9#2162
12Hulu-Med-4B41.4#2392
13HuluMed-7B36.5#2049
14MedGemma-4B24.2#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-image-quality · How It Works · Data refreshed daily, snapshot 2026-11-21.