EHRBench - Prognosis: leaderboard

EHRBench (2026) tests LLMs on clinical decision questions generated from real patient records and verified against biomedical knowledge bases, posed as multiple-choice questions with 4 to 6 options. This board is accuracy in percent on the prognosis-decision questions (predict a diagnosis at the next encounter); higher is better.

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 February 2027. 31 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.2 Instant60.59#304
2GPT-5 Chat58.46#449
3GPT-4.158.33#505
4GPT-5 Mini56.79#364
5GPT-4.1 Mini56.05#789
6Qwen 3 32B55.04#972
7Llama 3.3 70B Instruct54.44#1385
8Mistral Small 353.81#1796
9GLM-4 32B (0414)53.36#1505
10m1-32B-1k51.73#1240
11Qwen 2.5 32B51.54#1344
12EHRBench llama3-70B (checkpoint unspecified)50.2
13doctor-r1-8B49.98#1624
14Qwen 3 8B49.93#1873
15GPT-4.1 Nano49.39#2056

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ehrbench-prognosis · How It Works · Data refreshed daily, snapshot 2026-11-21.