Med-R2 - Joint QA: leaderboard

Accuracy (%) of medical vision-language models on the Med-R2 (2026) joint qa task, answering all reasoning levels for one image in a single prompt, with equal numbers of positive, negative and neutral adversarial question forms; higher is better.

Metric: Accuracy (%) on the Med-R2 Joint QA task (answering all reasoning levels for one image in a single prompt), 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
1GPT-5.2Thinking90.2#185
2Qwen 3 VL 8B75.3#917
3InternVL3.5-8B73.8#1604
4Qwen 3 VL 235B A22B72.5#485
5GPT-4o65.5#751
6Hulu-Med-14B63.1#1607
7Lingshu-7B60.2#1635
8Hulu-Med-4B58.4#2392
9Qwen 2.5 VL 7B Instruct57.3#1902
10HuaTuoGPT-Vision-34B54.8#1828
11HuatuoGPT-Vision-7B51#2162
12HuluMed-7B48.9#2049
13MedGemma-4B45.3#2557
14Janus-Pro-7B32.7#6857

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-joint-qa · How It Works · Data refreshed daily, snapshot 2026-11-21.