MedQ-Bench - Perception: leaderboard
Multiple-choice questions on medical image quality — whether a multimodal model can name a degradation, say which kind it is, and rate how severe it is across CT, MRI, histopathology, endoscopy and fundus images.
Metric: Overall Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 68.97 | #98 |
| 2 | GPT-4o | 64.79 | #343 |
| 3 | Grok 4 | 63.14 | #181 |
| 4 | Gemini 2.5 Pro | 61.88 | #152 |
| 5 | InternVL3-38B | 61 | #405 |
| 6 | Claude Sonnet 4 | 60.23 | #199 |
| 7 | InternVL3-8B | 60.08 | #620 |
| 8 | Mistral Medium 3 | 57.7 | #500 |
| 9 | Qwen 2.5 VL 7B | 54.71 | #643 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=medq-bench-perception · How It Works · Data refreshed daily, snapshot 2026-10-20.