Medmarks: leaderboard
Living medical-model benchmark spanning verifiable and open-ended clinical evaluation families with model/system comparisons.
Metric: Score (%). Source: medmarks.ai. Status: years away from saturation. 71 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 96.8 |
| 2 | Grok 4 | 95.07 |
| 3 | GPT-5.1 (Medium) | 94.91 |
| 4 | GPT-5.2 (Medium) | 94.69 |
| 5 | GLM-4.7 FP8 | 94.35 |
| 6 | Claude Sonnet 4.5 | 93.54 |
| 7 | Qwen 3 235B A22B (Thinking) | 93.13 |
| 8 | Qwen 3 Next 80B A3B (Thinking) | 92.14 |
| 9 | MiniMax-M2.1 | 91.84 |
| 10 | Qwen 3 Next 80B A3B Instruct | 91.61 |
| 11 | MiniMax-M2 | 91.34 |
| 12 | GPT-OSS-120B (High) | 90.77 |
| 13 | INTELLECT-3 | 90.65 |
| 14 | GPT-OSS-120B (Medium) | 90.19 |
| 15 | Qwen 3 VL 30B A3B (Thinking) | 90.11 |
Interactive version: theaggregate.ai/benchmark?slug=medmarks · How It Works · Data refreshed daily, snapshot 2026-09-05.