OpenAI Model Outperforms ER Doctors on Real Clinical Cases
Harvard researchers tested an OpenAI reasoning model on actual emergency room cases, not exam questions. The model diagnosed more accurately than the human doctors on those same cases. Nobody's getting replaced tomorrow, but it's reshaping what counts as a fast, reliable second opinion in the ER.
practitioners › Study published in Science, April 30, 2026, real ER cases and clinical vignettes, OpenAI reasoning model not publicly specified.
Science →