Medmarks - PubMedQA: leaderboard
The "PubMedQA" subtask of Medmarks. Living medical-model benchmark spanning verifiable and open-ended clinical evaluation families with model/system comparisons.
Metric: Score (%). Source: medmarks.ai. Saturation forecast: Around January 2027. 89 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 81.27 | #66 |
| 2 | Qwen 3.6 27B (Thinking) | 81.13 | #170 (Qwen 3.6 27B) |
| 3 | Qwen 3.5 27B (Thinking) | 80.93 | #187 (Qwen 3.5 27B) |
| 4 | Qwen 3.5 35B A3B (Thinking) | 80.6 | #256 (Qwen 3.5 35B A3B) |
| 5 | Qwen 3.6 35B A3B (Thinking) | 80.47 | #226 (Qwen 3.6 35B A3B) |
| 6 | Gemma 4 31B (IT) | 79.73 | #176 |
| 7 | Llama 3.3 70B Instruct | 79.33 | #534 |
| 8 | Qwen 3 VL 30B A3B (Thinking) | 78.53 | #351 (Qwen 3 VL 30B A3B) |
| 9 | Grok 4 | 78.2 | #181 |
| 10 | GPT-OSS-120B (Medium) | 78.13 | #341 (GPT-OSS-120B) |
| 11 | GPT-5.1 (Medium) | 77.67 | #135 (GPT-5.1) |
| 12 | Qwen 3 30B A3B (Thinking) | 77.47 | #493 (Qwen 3 30B A3B) |
| 13 | GPT-OSS-120B (Low) | 77.4 | #341 (GPT-OSS-120B) |
| 14 | GLM-4.7 FP8 | 77.4 | #209 |
| 15 | Qwen 3 235B A22B (Thinking) | 77.07 | #313 (Qwen 3 235B A22B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=medmarks-pubmedqa · How It Works · Data refreshed daily, snapshot 2026-10-20.