ChatGPT just beat doctors at medicine — and it's free
openai's gpt‑5.5 instant now outperforms physicians in blind evaluations on health accuracy and safety, bringing frontier medical intelligence to all free users.
openai released gpt‑5.5 instant, a model that achieved superior ratings compared to physician-written responses in head-to-head evaluations involving 3,500 reviews by a separate panel of doctors. the model demonstrated fewer failure modes than physicians, such as missing red flags or failing to seek necessary context, and matched the performance of openai's frontier thinking models on health-specific benchmarks like healthbench professional. monitoring of billions of production messages shows a 71% reduction in factuality issues over the last two months. this development extends high-level health intelligence to free users, signaling a significant step where general-purpose models outperform human experts in specific medical communication tasks.