PMC-VQA
Health & Medicine · 2023-05-15
PMC-VQA is a medical visual question-answering benchmark built from PubMed Central figures and captions, with four-option multiple-choice questions spanning many imaging modalities and diseases. Scores report multiple-choice accuracy on the test set.
Top models (higher is better)
| Model | Score |
|---|---|
| Qwen3.8-Max | 66.2 |
| Qwen3.5 397B A17B | 64.2 |
| Qwen3.7-Plus (source unspecified) | 63.4 |
| Kimi K2.5 | 63.3 |
| Qwen3.5 122B A10B | 63.3 |
| Qwen3.5-Omni-Plus | 62.7 |
| Gemini 3.1 Pro Preview | 62.5 |
| Qwen3.5-27B | 62.4 |
| Qwen3.5 Plus (2026-02-15) | 62.4 |
| Gemini 3 Pro Preview | 62.3 |
| GPT-5.6 Sol | 62.3 |
| Qwen3.5 35B A3B | 62.0 |
| Opus 4.5 | 59.9 |
| Opus 4.8 | 59.2 |
| GPT-5.2 | 58.9 |
| Qwen3.5-Omni-Flash | 58.7 |
| Qwen3.5-9B | 57.9 |
| Sonnet 4.5 | 55.9 |
| Qwen3.5-4B | 55.5 |
| Qwen3.5-2B | 54.0 |
| Gemini 2.5 Flash-Lite | 48.8 |
| Qwen3.5-0.8B | 45.5 |
| GPT-5 Nano | 37.8 |
| GPT-5 Mini | 36.3 |