MedQ-Bench - Comparison Reasoning: leaderboard
The same four judged criteria as the no-reference board, but the model must compare two medical images and justify which has the better acquisition quality.
Metric: Overall Score (0-7). Source: huggingface.co. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 6.34 | #98 |
| 2 | GPT-4o | 5.71 | #343 |
| 3 | Grok 4 | 5.66 | #181 |
| 4 | Gemini 2.5 Pro | 5.59 | #152 |
| 5 | InternVL3-8B | 5.53 | #620 |
| 6 | Claude Sonnet 4 | 5.33 | #199 |
| 7 | Mistral Medium 3 | 5.24 | #500 |
| 8 | InternVL3-38B | 5.14 | #405 |
| 9 | Qwen 2.5 VL 7B | 4.08 | #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-comparison-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.