Improving health intelligence in ChatGPT

OpenAI has announced that GPT-5.5 Instant brings substantial improvements to health and wellness responses in ChatGPT, which is used by more than 230 million people weekly. According to the company, GPT-5.5 Instant demonstrates enhanced reasoning, better context awareness, clearer communication, and improved recognition of when urgent care is necessary. On challenging evaluations including HealthBench and HealthBench Professional, the new model performs comparably to frontier Thinking models and significantly improves upon GPT-5.3 Instant. Both models are available to free users in ChatGPT. This progress is largely driven by a global network of more than 260 physicians across 60 countries, 49 languages, and 26 medical specialties. These medical professionals have reviewed more than 700,000 example responses to establish rubrics measuring accuracy, safety, clarity, completeness, and appropriate escalation. In evaluation comparisons, physicians rated GPT-5.5 Instant responses higher than older models and even physician-written responses in certain criteria, citing fewer failure modes such as missed red flags or lack of local healthcare context. Furthermore, privacy-preserving monitors tracking billions of weekly messages on production traffic indicate that the rate of responses with at least one flagged factuality issue in health has dropped by 71 percent over the last two months. OpenAI states that this physician-informed approach also supports its broader healthcare tools, including ChatGPT for Clinicians and OpenAI for Healthcare.