Medmarks - MedHallu Easy: leaderboard
The "MedHallu Easy" 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) | 86.5 | #66 |
| 2 | Gemma 4 31B (IT) | 84.12 | #176 |
| 3 | Claude Sonnet 4.5 | 83.83 | #142 |
| 4 | Qwen 3.6 27B (Thinking) | 82.67 | #170 (Qwen 3.6 27B) |
| 5 | Qwen 3.6 35B A3B (Thinking) | 82.67 | #226 (Qwen 3.6 35B A3B) |
| 6 | GPT-5.1 (Medium) | 80.07 | #135 (GPT-5.1) |
| 7 | Qwen 3.5 35B A3B (Thinking) | 77.38 | #256 (Qwen 3.5 35B A3B) |
| 8 | Qwen 3.5 27B (Thinking) | 76.47 | #187 (Qwen 3.5 27B) |
| 9 | Ling-flash-2.0 | 76.47 | #550 |
| 10 | Llama 3.3 70B Instruct | 74.31 | #534 |
| 11 | GPT-OSS-120B (High) | 73.45 | #341 (GPT-OSS-120B) |
| 12 | GLM-4.7 FP8 | 72.23 | #209 |
| 13 | INTELLECT-3 | 70.67 | #427 |
| 14 | Grok 4 | 70.41 | #181 |
| 15 | Qwen 3 235B A22B (Thinking) | 69.89 | #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-medhallu-easy · How It Works · Data refreshed daily, snapshot 2026-10-20.