HealthAgentBench - EHR Event Modelling: leaderboard
HealthAgentBench (2026) evaluates coding agents on healthcare tasks in terminal environments over real clinical data, at xhigh reasoning effort with web browsing disabled. This board is the task success rate in percent over three attempts at building clinical event prediction models on longitudinal EHR data, a trial succeeding when test AUROC reaches a count-feature gradient-boosting baseline; higher is better.
Metric: Longitudinal EHR clinical event prediction adapted from EHRSHOT (6 tasks; success when test AUROC reaches the count plus gradient-boosting baseline): task success rate (%) over three attempts per task, a trial succeeding only when it meets the category success criterion; terminal environments in the Harbor framework, xhigh reasoning effort, web browsing disabled. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 (xHigh) | 78 | #79 (Claude Opus 4.7) |
| 2 | GPT-5.5 (xHigh) | 72 | #49 (GPT-5.5) |
| 3 | Claude Opus 4.8 (xHigh) | 67 | #56 (Claude Opus 4.8) |
| 4 | GPT-5.4 (xHigh) | 61 | #123 (GPT-5.4) |
| 5 | GPT-5.3 Codex (xHigh) | 61 | #116 (GPT-5.3 Codex) |
| 6 | GPT-5.4 Mini (xHigh) | 39 | #389 (GPT-5.4 Mini) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=healthagentbench-ehr-event-modelling · How It Works · Data refreshed daily, snapshot 2026-11-13.