MedXpertQA Text
Health & Medicine · 2025-01-30
MedXpertQA Text is the full 2,450-question text-only split of MedXpertQA, expert-level medical examination questions drawn from specialty board and licensing exams across 17 specialties and 11 body systems. Each question has ten answer options and is scored by answer accuracy. Higher is better.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.1 Pro Preview | 71.5 |
| GPT-5.4 | 59.6 |
| o3 | 54.6 |
| Muse Spark | 52.6 |
| Opus 4.6 | 52.1 |
| Grok 4.20 | 50.2 |
| Gemini 2.5 Pro | 43.1 |
| Gemini 2.5 Flash | 36.2 |
| MedGemma 27B Text IT | 25.7 |
| MedGemma 27B IT | 23.7 |
| MedGemma 1.5 4B IT | 16.4 |
| Gemma 3 27B IT | 15.7 |
| MedGemma 4B IT | 14.2 |
| Gemma 3 4B IT | 11.6 |