Study finds OpenAI large language model outperforms doctors in diagnostic and clinical reasoning tasks

Researchers from Boston published a paper in Science describing a series of experiments that tested an OpenAI large language model on real‑world diagnostic and clinical reasoning cases drawn from an emergency department.

The model performed better than practicing physicians on case‑based evaluations, achieving higher accuracy in identifying conditions and suggesting appropriate next steps.

The authors highlight the need for rigorous clinical trials before such tools are deployed widely, warning that marketing hype could outpace solid evidence.

They see the results as a step toward integrating AI into clinical decision support, but stress that safety, validation, and regulatory oversight remain essential.