EHRBench - Prognosis: leaderboard
Metric: Accuracy (%) on the prognosis-decision questions (predict a diagnosis at the next encounter), 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 August 2027. 31 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 Instant | 60.59 |
| 2 | GPT-5 Chat | 58.46 |
| 3 | GPT-4.1 | 58.33 |
| 4 | GPT-5 Mini | 56.79 |
| 5 | GPT-4.1 Mini | 56.05 |
| 6 | Qwen 3 32B | 55.04 |
| 7 | Llama 3.3 70B Instruct | 54.44 |
| 8 | Mistral Small 3 | 53.81 |
| 9 | GLM-4 32B (0414) | 53.36 |
| 10 | Qwen 2.5 32B | 51.54 |
| 11 | Qwen 3 8B | 49.93 |
| 12 | GPT-4.1 Nano | 49.39 |
| 13 | Qwen 3 4B | 48.76 |
| 14 | GLM-4 9B (0414) | 47.9 |
| 15 | Yi 1.5 34B | 47.86 |
Interactive version: theaggregate.ai/benchmark?slug=ehrbench-prognosis · How It Works · Data refreshed daily, snapshot 2026-10-07.