HealthAgentBench - X-ray Report Correction: 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 correcting chest X-ray reports, a trial succeeding when at least three of five judge votes find no clinically significant error; higher is better.
Metric: Chest X-ray report correction (10 tasks; success when at least 3 of 5 judge votes find zero clinically significant errors): 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 2028. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (xHigh) | 40 | #123 (GPT-5.4) |
| 2 | GPT-5.5 (xHigh) | 33 | #49 (GPT-5.5) |
| 3 | GPT-5.3 Codex (xHigh) | 20 | #116 (GPT-5.3 Codex) |
| 4 | Claude Opus 4.8 (xHigh) | 17 | #56 (Claude Opus 4.8) |
| 5 | Claude Opus 4.7 (xHigh) | 13 | #79 (Claude Opus 4.7) |
| 6 | GPT-5.4 Mini (xHigh) | 10 | #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-x-ray-report-correction · How It Works · Data refreshed daily, snapshot 2026-11-13.