AutoMedBench - Segmentation: leaderboard

Metric: Overall run score (0-100) on the segmentation track (8 CT, MRI and ultrasound tasks): the mean of the held-out task score (macro Dice, 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 February 2027. 6 models tracked.

Top models

#ModelScore
1Claude Opus 4.667.2
2GPT-5.459.4
3GLM-558
4Gemini 3.1 Pro (Preview)54.7
5MiniMax-M2.546.5
6Qwen 3.5 397B A17B42.8

Interactive version: theaggregate.ai/benchmark?slug=automedbench-segmentation · How It Works · Data refreshed daily, snapshot 2026-09-29.