MedMosaic - Long-Form Consultation MCQ: leaderboard

Metric: Accuracy (%; multiple choice, extracted answer letter against the answer key; 1,074 questions on clinical conversations longer than three minutes). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash73.6
2Gemini 2.5 Pro67.4
3GPT-4o Audio58.7
4Qwen2.5-Omni-7B39
5Phi-4 Multimodal Instruct24.7
6Gemma 3n E4B (IT)24.5

Interactive version: theaggregate.ai/benchmark?slug=medmosaic-long-form-consultation-mcq · How It Works · Data refreshed daily, snapshot 2026-09-26.