OpenAI Boosts ChatGPT Health AI

OpenAI is rolling out significant upgrades to ChatGPT's health capabilities with the introduction of GPT-5.5 Instant. This update aims to provide more reliable and understandable health information to its massive user base, which sees over 230 million weekly interactions for health and wellness queries.

The company claims that GPT-5.5 Instant now performs comparably to its most advanced 'Thinking' models on challenging health evaluations. Crucially, these improvements are available to all free ChatGPT users, democratizing access to what OpenAI terms 'frontier health intelligence'.

Physician-Led Evaluation Drives Progress

This leap in performance is attributed to advancements in OpenAI's underlying models and a rigorous physician-led evaluation process. A global network of over 260 physicians across 60 countries contributes to defining and measuring 'good' health responses.

Physicians review example model outputs, identifying areas for improvement in accuracy, clarity, context awareness, and appropriate escalation of care. This feedback loop has led to reviews of over 700,000 model responses, ensuring ChatGPT's health advice is safer and more useful.

Measuring Health Intelligence

OpenAI utilizes health-specific evaluations like HealthBench and HealthBench Professional to quantify progress. These frameworks assess qualities such as accuracy, safety, and communication, using physician-written rubrics against realistic health scenarios. The latest GPT-5.5 Instant model shows substantial gains over GPT-5.3 Instant on these metrics.

In a comparative study, physicians also rated GPT-5.5 Instant responses higher than both older AI models and human-written responses for representative health conversations. The AI demonstrated fewer instances of failing to tailor to local context, missing critical red flags, or neglecting to seek necessary user clarification.

Real-World Impact and Future Goals

Beyond structured evaluations, OpenAI monitors production traffic for factuality issues. They report a 71% reduction in flagged health responses over the past two months, indicating a significant real-world improvement in reliability.

This work on GPT-5.5 Instant health intelligence complements OpenAI's broader healthcare initiatives, including tools designed for medical professionals. The company views improving human health as a primary impact of advanced AI.