OpenAI's GPT-5.6 AI model family beats human physicians in a blind clinical evaluation of health responses

"physicians found fewer flaws in GPT-5.6 responses than physician-written responses." https://twitter.com/thekaransinghal/status/2075689779937833302

I’m very proud of how small the gap is between our free and paid models at healthcare! https://twitter.com/thekaransinghal/status/2075689779937833302

@thekaransinghal Looks encouraging, Karan. Is there a preprint or a white paper to review with the methods and results?

Combined views