PatientAgentBench - Triage Quality: leaderboard

Metric: Mean triage quality rubric score (1-5); 1,200 synthetic patient conversations of up to 15 turns with a simulated patient, the model inside the benchmark's baseline ReAct agent harness with 15 healthcare sandbox tools; jury of Claude Opus 4.8 and GPT-5.5 judges scoring clinician-grounded rubrics on a 1 to 5 scale; higher is better. Source: arxiv.org. Saturation forecast: Around May 2027. 10 models tracked.

Top models

#ModelScore
1Claude Sonnet 53.77
2Claude Opus 4.83.7
3GPT-5.53.49
4GPT-5.43.45
5Gemini 3 Flash3.03
6Claude Haiku 4.52.8
7GPT-OSS-120B2.78
8Gemini 3.1 Pro (Preview)2.66
9Qwen 3 235B A22B2.55

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