ClinTraceBench - Cross-Patient Comparison: leaderboard

Metric: Accuracy (%; task T6, cross-patient comparison: number of diagnoses, maximum lab value and lab delta across patients, 700 questions; full-context strategy: the whole multi-visit patient-clinician dialogue derived from MIMIC-IV records is given verbatim, models called through OpenRouter). Source: arxiv.org. Saturation forecast: Around 2029. 4 models tracked.

Top models

#ModelScore
1DeepSeek V3.154.3
2Claude Sonnet 4.652.3
3Claude Haiku 4.551
4GPT-4o Mini49.7

Interactive version: theaggregate.ai/benchmark?slug=clintracebench-cross-patient-comparison · How It Works · Data refreshed daily, snapshot 2026-09-26.