MedFlowBench (Pathology) - Breast WSI Strict: leaderboard

Metric: Strict score (0-1, shown times 100) on BRACS breast whole-slide images (113 slides: seven-class diagnosis): the diagnosis must be right and, for non-normal slides, a returned coordinate must fall inside an annotated region of the true class; the model drives QuPath through the MedOpenClaw runtime with viewer-native actions only (slide loading, panning, zooming, viewport capture), one tool call per round within a 20-round budget, returning an answer with whole-slide coordinates as evidence (Track A); higher is better. Source: arxiv.org. Saturation forecast: Around August 2028. 4 models tracked.

Top models

No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.

Interactive version: theaggregate.ai/benchmark?slug=medflowbench-pathology-breast-wsi-strict · How It Works · Data refreshed daily, snapshot 2026-10-11.