MMRareBench - Treatment Planning: leaderboard

Metric: Treatment Planning track (T2) score: a stage-wise treatment plan given the working diagnosis, case narrative and images, graded 0-100 by a Qwen3-VL-235B judge with a strict track-specific rubric that awards credit only when a criterion is fully met (six binary dimensions averaged); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 23 models tracked.

Top models

#ModelScore
1GPT-549.2
2Qwen 3 VL 235B A22B Instruct35.9
3Claude Sonnet 4.535.8
4Claude Haiku 4.534.2
5Gemini 3 Flash (Preview)29.1
6Gemini 2.5 Pro26.7
7GLM-4.6V20.5
8Qwen 3 VL 30B A3B Instruct20.1
9Gemini 2.5 Flash17.1
10MedGemma-27B-IT15.4
11Lingshu-32B14.2
12Qwen 2.5 VL 32B Instruct12.8
13GPT-4o12.7
14Qwen 3 VL 8B Instruct10.8
15Qwen 2.5 VL 72B Instruct8.7

Interactive version: theaggregate.ai/benchmark?slug=mmrarebench-treatment-planning · How It Works · Data refreshed daily, snapshot 2026-10-07.