MedRealMM (Physician Subset): leaderboard
MedRealMM (JD Health, 2026) turns de-identified multimodal consultations from a Chinese internet hospital into next-physician-response tasks with patient images and physician-refined case rubrics, graded per criterion by Gemini-3-Pro-Preview. This board is the 200-case subset used to compare models with deliberating physicians; score is the case-mean rubric score in percent, higher is better.
Metric: Rubric score (%; case-level mean of the physician-refined case rubric, positive and negative criteria weighted and clipped to 0-1, graded per criterion by Gemini-3-Pro-Preview) on the 200-case subset of real Chinese online consultations with patient images used for the deliberative physician comparison. Source: arxiv.org. Saturation forecast: Around March 2028. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 50.66 | #79 |
| 2 | Gemini 3.5 Flash | 44.93 | #101 |
| 3 | Kimi K2.6 | 41.78 | #178 |
| 4 | Qwen 3.6 27B | 41.2 | #341 |
| 5 | GPT-5.5 | 38.53 | #49 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medrealmm-physician-subset · How It Works · Data refreshed daily, snapshot 2026-11-13.