Medmarks - PubMedQA: leaderboard
Metric: Score (%). Source: medmarks.ai. 71 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 81.27 |
| 2 | Llama 3.3 70B Instruct | 79.33 |
| 3 | Qwen 3 VL 30B A3B (Thinking) | 78.53 |
| 4 | Grok 4 | 78.2 |
| 5 | GPT-OSS-120B (Medium) | 78.13 |
| 6 | GPT-5.1 (Medium) | 77.67 |
| 7 | GPT-OSS-120B (Low) | 77.4 |
| 8 | GLM-4.7 FP8 | 77.4 |
| 9 | Qwen 3 235B A22B (Thinking) | 77.07 |
| 10 | Qwen 3 Next 80B A3B Instruct | 76.87 |
| 11 | GPT-OSS-120B (High) | 76.67 |
| 12 | MiniMax-M2 | 76.6 |
| 13 | Ling-flash-2.0 | 76.47 |
| 14 | GPT-OSS-20B (High) | 76.33 |
| 15 | Claude Sonnet 4.5 | 76.27 |
Interactive version: theaggregate.ai/benchmark?slug=medmarks-pubmedqa · How It Works · Data refreshed daily, snapshot 2026-09-05.