AutoMedBench - Report Generation: leaderboard
Metric: Overall run score (0-100) on the report generation track (5 chest X-ray and pathology report tasks): the mean of the held-out task score (mean of BLEU, METEOR, ROUGE-L, F1RadGraph and micro precision, recall and F1, scaled to 0-100) and the workflow agentic score (Plan, Setup, Validate, Inference and Submit stages weighted 25/15/35/15/10, LLM judge for the first three); each model is the single agent of the benchmark harness (shared code-execution interface, system prompt, tool schema and default decoding; no vendor agent framework), runs averaged over the Lite and Standard tiers; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 55.8 |
| 2 | GLM-5 | 48.6 |
| 3 | Gemini 3.1 Pro (Preview) | 47.7 |
| 4 | GPT-5.4 | 40.8 |
| 5 | Qwen 3.5 397B A17B | 38.7 |
| 6 | MiniMax-M2.5 | 28.9 |
Interactive version: theaggregate.ai/benchmark?slug=automedbench-report-generation · How It Works · Data refreshed daily, snapshot 2026-09-29.