HealthAgentBench - Clinical Trial Matching: 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 matching patients to clinical trials from free-text protocols, a trial succeeding when every eligible trial is among the 50 most confident picks; higher is better.
Metric: Clinical trial matching over about 400 free-text protocols (9 tasks; success when every eligible trial is among the 50 most confident picks): 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 | GPT-5.5 (xHigh) | 52 | #49 (GPT-5.5) |
| 2 | Claude Opus 4.8 (xHigh) | 48 | #56 (Claude Opus 4.8) |
| 3 | Claude Opus 4.7 (xHigh) | 26 | #79 (Claude Opus 4.7) |
| 4 | GPT-5.4 (xHigh) | 18 | #123 (GPT-5.4) |
| 5 | GPT-5.4 Mini (xHigh) | 18 | #389 (GPT-5.4 Mini) |
| 6 | GPT-5.3 Codex (xHigh) | 18 | #116 (GPT-5.3 Codex) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=healthagentbench-clinical-trial-matching · How It Works · Data refreshed daily, snapshot 2026-11-13.