CURE (Clinical Diagnosis, Physician References) - Top-3: leaderboard
Metric: Open-ended diagnosis, Hit@3 in percent: share of cases whose reference diagnosis is among the model's top 3 ranked diagnoses (matched by a DeepSeek-V3.1 judge, kappa 0.86 against a physician on 50 cases), on CURE's 500 test cases (Eurorad 2025 radiology cases: clinical history, a mean of 8 images, radiologist findings with diagnostic statements removed), zero-shot, given the clinical history, images and the abstracts of the physician-cited references; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 12B (IT) | 90 | #655 |
| 2 | GPT-5.2 | 81.8 | #105 |
| 3 | Gemma 3 4B (IT) | 81.2 | #971 |
| 4 | Qwen 3 VL 32B Instruct | 80.4 | #276 |
| 5 | Gemini 3 Pro (Preview) | 80 | #64 |
| 6 | Qwen 3 VL 235B A22B Instruct | 79.4 | #264 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 79.4 | #228 (Qwen 3 VL 235B A22B) |
| 8 | Kimi K2.5 | 77.4 | #139 |
| 9 | GLM-4.6V | 77 | #309 |
| 10 | Qwen 3 VL 8B Instruct | 76.2 | #401 |
| 11 | Qwen 3 VL 30B A3B Instruct | 76 | #365 |
| 12 | Qwen 3 VL 32B (Thinking) | 75.8 | #287 (Qwen 3 VL 32B) |
| 13 | MedGemma-4B-IT | 74.6 | #842 |
| 14 | Qwen 3 VL 8B (Thinking) | 72.2 | |
| 15 | Qwen 3 VL 30B A3B (Thinking) | 72 | #338 (Qwen 3 VL 30B A3B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=cure-clinical-diagnosis-physician-references-top-3 · How It Works · Data refreshed daily, snapshot 2026-10-11.