MedMosaic - Speech and Sound MCQ: leaderboard
MedMosaic (Rajgarhia et al., 2026): 10,884 multiple-choice questions on synthetic doctor-patient conversations with inserted coughs, wheezes and other clinical sounds. Accuracy in percent (higher is better).
Metric: Accuracy (%; multiple choice, extracted answer letter against the answer key; 10,884 questions on synthetic MTS-Dialog consultations with inserted clinical sounds). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 68.3 | #262 |
| 2 | Gemini 2.5 Flash | 60.2 | #452 |
| 3 | Gemma 3n E4B (IT) | 33.3 | #2220 |
| 4 | Qwen2.5-Omni-7B | 32.5 | #1797 |
| 5 | Phi-4 Multimodal Instruct | 26.6 | #3914 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=medmosaic-speech-and-sound-mcq · How It Works · Data refreshed daily, snapshot 2026-11-12.