Atlas

Benchmarks

← All benchmarks

MedQA

Health & Medicine · 2020-09-28

MedQA evaluates medical question-answering ability on medical exam-style questions. Scores are usually reported as accuracy percentages.

Top models (higher is better)

ModelScore
o393.3
Gemini 2.5 Pro92.6
Gemini 2.5 Flash92.0
MedGemma 27B Text IT87.7
MedGemma 27B IT85.3
Gemma 3 27B IT74.9
MedGemma 1.5 4B IT69.1
MedGemma 4B IT64.4
Gemma 3 4B IT50.7
Loading Atlas data…