MedGUIDE: leaderboard
MedGUIDE evaluates whether LLMs can make guideline-consistent clinical decisions. It is constructed from 55 curated NCCN decision trees across 17 cancer types and uses LLM-generated clinical scenarios filtered through expert-labeled reward models and LLM-as-judge ensembles to produce 7,747 high-quality multiple-choice diagnostic questions. (Source: benchmarklist.com, self-reported.)
Metric: Weighted Accuracy (self-reported). Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 62.54 | #248 |
| 2 | O1 | 60.49 | #183 |
| 3 | O4 Mini | 58.9 | #180 |
| 4 | DeepSeek R1 | 58.12 | #251 |
| 5 | DeepSeek V3 | 57.7 | #321 |
| 6 | Claude 3.7 Sonnet (20250219) | 55.62 | #201 |
| 7 | Llama 3.1 70B Instruct | 54.31 | #561 |
| 8 | Qwen 3 32B | 53.66 | #443 |
| 9 | Qwen 3 14B | 53.54 | #537 |
| 10 | GPT-4o Mini | 53.42 | #599 |
| 11 | Qwen 2.5 7B Instruct | 50.07 | #862 |
| 12 | Mixtral 8x7B Instruct (v0.1) | 49.15 | #1073 |
| 13 | Mistral 7B Instruct (v0.3) | 48.13 | #1274 |
| 14 | Claude 3.5 Haiku (20241022) | 47.55 | #588 |
| 15 | Qwen 3 8B | 47.33 | #674 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=medguide · How It Works · Data refreshed daily, snapshot 2026-10-20.