OpenAI says its GPT-5.5 Instant model is making ChatGPT more useful for health-related questions, with improvements in how it handles urgency, uncertainty and follow-up context. The company said the update is now available to free users, broadening access to the latest health-focused gains.
OpenAI said health is among the most common uses of ChatGPT, with more than 230 million people each week turning to the product for questions about symptoms, lab results, appointments, insurance and general wellness. The company said GPT-5.5 Instant marks a significant step forward in those conversations, especially when users need help understanding whether a situation may require prompt care.
According to OpenAI, the new model is better at spotting when urgent medical attention may be needed, asking for relevant details, acknowledging uncertainty and translating complex information into simpler language. The company said its internal health evaluations show GPT-5.5 Instant performing at a level comparable to its frontier Thinking models on the most difficult health tests.
The update also reflects OpenAI’s physician-led evaluation process. The company said it works with a global network of doctors who help define what strong medical responses should look like in practical situations. Those physicians review example answers, flag weaknesses and help build rubrics for measuring performance.
OpenAI said it uses health-specific benchmarks such as HealthBench and HealthBench Professional to test qualities including accuracy, safety, communication, completeness, context awareness and whether the model escalates care appropriately. In one comparison, physicians reviewed model answers and physician-written answers to representative health questions, with OpenAI saying the GPT-5.5 Instant responses were rated higher across the reviewed criteria.
The company also said physicians found fewer failure modes in GPT-5.5 Instant than in older models and physician-written responses. Examples cited included better tailoring to local health context, fewer missed warning signs and more frequent requests for additional information when needed.
OpenAI pointed to a separate analysis of live traffic as another sign of improvement. The company said its privacy-preserving monitoring of health-related production messages found that the rate of responses with at least one flagged factuality issue has dropped by 71% over the last two months, based on billions of messages a week.
To illustrate the changes, OpenAI compared answers to a question about why a doctor might order an MRI before a steroid injection for sciatica. The company said newer responses were more structured and more careful, explaining that imaging can help confirm the cause of pain, identify the right injection site, check for red flags and determine whether another treatment might be better.
OpenAI said the health work is tied to broader product efforts aimed at clinicians and healthcare organizations, including ChatGPT for Clinicians and OpenAI for Healthcare. The company framed the improvements as part of a longer-term goal of making ChatGPT more accurate and more useful in situations that involve personal health decisions.
The company said it sees better health support as one of the clearest possible benefits of more capable AI systems. For now, GPT-5.5 Instant is the latest example of how OpenAI is trying to make those gains available to more users, including those on the free tier of ChatGPT.