EHRBench - MIMIC-III: leaderboard
Metric: Accuracy (%) on the questions built from MIMIC-III records, EHR-grounded multiple-choice clinical decision questions (4, 5 and 6 options) built from MIMIC-III, MIMIC-IV and PROMOTE encounter trajectories with knowledge-base verification, JSON-constrained answers, deterministic decoding; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 31 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 Instant | 71.5 |
| 2 | GPT-4.1 | 70.59 |
| 3 | GPT-5 Chat | 70.16 |
| 4 | GPT-5 Mini | 69.08 |
| 5 | Llama 3.3 70B Instruct | 68.74 |
| 6 | GPT-4.1 Mini | 68.58 |
| 7 | Qwen 3 32B | 68.48 |
| 8 | Mistral Small 3 | 67 |
| 9 | GLM-4 32B (0414) | 66.27 |
| 10 | Qwen 2.5 32B | 65.22 |
| 11 | Qwen 3 8B | 62.26 |
| 12 | Qwen 3 4B | 62.01 |
| 13 | GLM-4 9B (0414) | 61.92 |
| 14 | GPT-4.1 Nano | 61.59 |
| 15 | Yi 1.5 34B | 60.72 |
Interactive version: theaggregate.ai/benchmark?slug=ehrbench-mimic-iii · How It Works · Data refreshed daily, snapshot 2026-10-07.