Science study: OpenAI's o1 beat physicians at diagnosis on real ER cases

Researchers at Harvard Medical School and Beth Israel Deaconess reported in Science that OpenAI's o1 reasoning model outperformed physician baselines across clinical-reasoning benchmarks and, on 76 real emergency-department cases using raw health-record data, named the exact or a very close diagnosis more often than attending physicians. The authors said prospective clinical trials are needed before autonomous use.