MMRareBench - Diagnosis F1: leaderboard
Metric: Diagnosis track (T1) token-level F1 (%, SQuAD-style normalization, maximum over the gold diagnosis aliases) of the predicted primary diagnosis from a masked rare-disease case narrative and its medical images; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 84.7 |
| 2 | Gemini 2.5 Pro | 80.1 |
| 3 | GPT-5 | 76.1 |
| 4 | Claude Sonnet 4.5 | 70.8 |
| 5 | GPT-4o | 70 |
| 6 | Gemini 2.5 Flash | 70 |
| 7 | Qwen 3 VL 235B A22B Instruct | 67.3 |
| 8 | Qwen 2.5 VL 72B Instruct | 61 |
| 9 | Qwen 3 VL 30B A3B Instruct | 56.3 |
| 10 | Lingshu-32B | 53.3 |
| 11 | Claude Haiku 4.5 | 51.8 |
| 12 | MedGemma-27B-IT | 51.3 |
| 13 | Qwen 2.5 VL 32B Instruct | 49.6 |
| 14 | Qwen 3 VL 8B Instruct | 46.7 |
| 15 | Qwen 2.5 VL 7B Instruct | 41 |
Interactive version: theaggregate.ai/benchmark?slug=mmrarebench-diagnosis-f1 · How It Works · Data refreshed daily, snapshot 2026-10-07.