Skip to main content
BenchmarkLLMsv1.0

MedQA

by Jin et al. / UC San Diego · open-source · Last verified 2026-03-17

MedQA tests medical knowledge using free-form multiple-choice questions drawn from the US Medical Licensing Examination (USMLE). It evaluates whether language models can reason through complex clinical scenarios requiring deep biomedical knowledge.

https://github.com/jind11/MedQA
B+
B+Good
Adoption: AQuality: A+Freshness: B+Citations: AEngagement: F

Specifications

License
Apache-2.0
Pricing
open-source
Capabilities
evaluation, benchmarking, medical-reasoning
Integrations
Use Cases
model-evaluation, medical-ai, clinical-nlp
API Available
No
Evaluated Models
gpt-4o, claude-opus-4, gemini-2-5-pro, meditron-70b, llama-3-70b
Metrics
accuracy, pass-rate
Methodology
Models are presented with four-option multiple-choice clinical vignettes from USMLE Steps 1-3. Accuracy is measured on the 1,273-question test split. No chain-of-thought is required by default.
Last Run
2026-02-15
Tags
medical, qa, clinical, multiple-choice, usmle
Added
2026-03-17
Completeness
100%

Index Score

72.8
Adoption
82
Quality
90
Freshness
74
Citations
88
Engagement
0

Explore the full AI ecosystem on Agents as a Service