OpenAI said a physician panel rated its new GPT-5.5 Instant model higher than responses written by physicians on a set of real-world health questions.
The AI developer said more than 230 million people use ChatGPT each week for health and wellness questions, including interpreting lab results, preparing for appointments and navigating insurance. GPT-5.5 Instant, released in May, improves on the March 2026 GPT-5.3 Instant in recognizing when urgent care may be needed, asking for relevant context and explaining uncertainty without overstating confidence, OpenAI said June 18.
A panel of physicians reviewed model outputs against responses physicians wrote with unlimited time and internet access across 3,500 reviewed responses. The physicians rated GPT-5.5 Instant higher than older models and the physician-written answers on accuracy, communication, completeness and health decision helpfulness, the company said.
OpenAI also reported that, based on privacy-preserving monitors of production traffic spanning billions of weekly messages, the rate of health responses flagged for a possible factuality issue has fallen 71% over the past two months.
