OpenAI reported on June 18, 2026 that GPT‑5.5 Instant has made measurable improvements in how ChatGPT handles health and wellness queries. The company says the model better recognizes when urgent care may be needed, requests relevant context from users, explains uncertainty more carefully, and renders complex information more clearly.
Evaluation results and physician review
According to OpenAI, GPT‑5.5 Instant performs at a level comparable to its frontier Thinking models on an aggregate of health evaluations, including HealthBench and HealthBench Professional. The company contrasted GPT‑5.5 Instant with earlier releases such as GPT‑5.3 Instant (released March 2026) and stated the new Instant variant was released in May 2026.
OpenAI described two evaluation approaches. In one, physicians wrote answers with internet access and unlimited time; a separate physician panel then compared those responses to model outputs across 3,500 reviewed examples, judging accuracy, communication, completeness, instruction following, and usefulness for health decisions. In that review, OpenAI reported that GPT‑5.5 Instant responses were rated higher than physician-written and older model responses on several criteria.
The company also uses production monitoring of billions of messages per week to track factuality. OpenAI reported a 71% decline, over two months, in the rate of health responses flagged with at least one factuality issue in production traffic.
How medical expertise was incorporated
OpenAI said it works with a global network of more than 260 physicians across 60 countries, 49 languages, and 26 medical specialties. These physicians have reviewed over 700,000 example model responses and helped create rubrics used in HealthBench evaluations to assess accuracy, safety, clarity, completeness, and appropriate escalation.
OpenAI also noted related tools such as ChatGPT for Clinicians and OpenAI for Healthcare that are part of its broader health work. GPT‑5.5 Instant is available to all free users in ChatGPT, subject to usage limits, and OpenAI reported these evaluation results and deployment details on June 18, 2026.
Original source: OpenAI News