MMTClinic - Task 1 (MCQ): leaderboard
Multimodal, multilingual clinical time-series question answering: Task 1 multiple-choice accuracy with text plus time-series input, averaged over languages. Higher is better. Source: arxiv.org/abs/2609.04842.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 66 | #262 |
| 2 | QwQ-32B | 52.37 | #422 |
| 3 | Qwen 3 235B A22B | 52.1 | #323 |
| 4 | GPT-4.1 Nano | 45.27 | #730 |
| 5 | Qwen 3 30B A3B | 43.77 | #496 |
| 6 | Llama 3.3 70B | 31.12 | #578 |
| 7 | Mistral 7B Instruct | 28.02 | #1350 |
| 8 | Gemini 2.0 Flash | 27.4 | #351 |
| 9 | Qwen 2.5 7B | 6.97 | #895 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmtclinic-task-1-mcq · How It Works · Data refreshed daily, snapshot 2026-11-05.