MedXpertQA: leaderboard
4,460 medical board-exam questions across 17 specialties and 11 body systems from Tsinghua University (2025): 2,450 text items with 10 options and 2,005 image-based items with 5 options; accuracy.
Metric: Score (mixed sources). Source: arxiv.org. Source: benchmarklist.com. Source: llm-stats.com. Saturation forecast: Around October 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 80.7 | #58 |
| 2 | Muse Spark | 78.4 | #64 |
| 3 | GPT-5.4 (xHigh) | 77.3 | #79 (GPT-5.4) |
| 4 | Qwen 3.7 Plus | 71 | #92 |
| 5 | Qwen 3.6 Plus | 68.7 | #120 |
| 6 | Qwen 3.5 122B A10B | 67.3 | #173 |
| 7 | Claude Opus 4.6 (Max) | 64.4 | #63 (Claude Opus 4.6) |
| 8 | Qwen 3.5 27B | 62.4 | #187 |
| 9 | Qwen 3.5 35B A3B | 61.4 | #256 |
| 10 | Gemma 4 31B | 61.3 | #200 |
| 11 | Gemma 4 26B A4B | 58.1 | #358 |
| 12 | Qwen 3.5 9B | 49.9 | #379 |
| 13 | O1 | 49.89 | #183 |
| 14 | DiffusionGemma 26B-A4B | 49 | #698 |
| 15 | Gemma 4 12B | 48.7 | #535 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=medxpertqa · How It Works · Data refreshed daily, snapshot 2026-10-20.