Open Medical LLM - MedQA (USMLE): leaderboard
MedQA (USMLE) subset: 4-option medical QA from the US Medical Licensing Examination, testing clinical reasoning and medical knowledge.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 193 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 | 78.8 | #438 |
| 2 | Llama 3 8B Slerp Biomed Chat Chinese | 63.55 | #1215 |
| 3 | NeuralLLaMa-3-8B-DT-v0.1 | 61.67 | #923 |
| 4 | Daredevil-8B-abliterated | 61.27 | #846 |
| 5 | ChimeraLlama-3-8B-v3 | 61.27 | #1055 |
| 6 | NeuralLLaMa-3-8B-ORPO-v0.3 | 61.19 | #1174 |
| 7 | Meta-Llama-3-8B-Instruct-abliterated-v3 | 61.04 | #1208 |
| 8 | Llama-3-8B-Instruct-abliterated-v2 | 60.96 | #1136 |
| 9 | Barcenas-Llama3-8B-ORPO | 60.88 | #949 |
| 10 | Daredevil-8B-abliterated-dpomix | 60.88 | #838 |
| 11 | llama-3-merged-linear | 60.8 | #887 |
| 12 | Llama 3 8B Instruct | 60.72 | #1129 |
| 13 | Llama 3 8B Ita | 60.72 | #926 |
| 14 | Llama-3-8B-Instruct-abliterated-dpomix | 60.64 | #1049 |
| 15 | Llama-3-SauerkrautLM-8B-Instruct | 60.57 | #895 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-medical-llm-medqa-usmle · How It Works · Data refreshed daily, snapshot 2026-10-20.