Frontier general-purpose large models scored higher than the physician control group in expert evaluations, particularly excelling in medical advice, treatment principles, and certain diagnostic tasks.