MIMIC-CDM - Appendicitis: leaderboard
MIMIC-CDM restricted to appendicitis cases: the model reads a real de-identified admission’s history, labs and imaging reports and has to reach the diagnosis. Accuracy on that pathology alone.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B (IT) | 89.9 | #523 |
| 2 | Gemma 3 12B (IT) | 86.1 | #664 |
| 3 | Gemma 2 27B | 85.9 | #806 |
| 4 | Llama 3.1 70B Instruct | 84.7 | #561 |
| 5 | Llama 3 70B Instruct | 84.6 | #634 |
| 6 | Llama 3.3 70B Instruct | 84.2 | #534 |
| 7 | DeepSeek R1 Distill Llama 70B | 78.8 | #594 |
| 8 | MedGemma-4B-IT | 67.7 | #853 |
| 9 | Gemma 3 4B (IT) | 64.3 | #992 |
| 10 | Gemma 3 1B (IT) | 45.2 | #1504 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mimic-cdm-appendicitis · How It Works · Data refreshed daily, snapshot 2026-10-20.