OpenAI Cuts ChatGPT Health Factuality Flags 71% With GPT-5.5 Instant
Updated
Updated · OpenAI · Jun 18
OpenAI Cuts ChatGPT Health Factuality Flags 71% With GPT-5.5 Instant
3 articles · Updated · OpenAI · Jun 18
Summary
71% fewer flagged factuality issues were recorded in ChatGPT health responses over the last two months, based on privacy-preserving monitoring of billions of weekly messages, as OpenAI rolled out GPT-5.5 Instant.
GPT-5.5 Instant also improved on urgent-care recognition, asking for missing context, explaining uncertainty and simplifying complex medical information; OpenAI said it now performs on its toughest health evaluations at a level comparable to frontier Thinking models.
230 million people use ChatGPT for health and wellness questions each week, and OpenAI said making GPT-5.5 Instant available to free users broadens access to those gains.
More than 260 physicians across 60 countries, 49 languages and 26 specialties help shape the system, and they have reviewed over 700,000 example responses to define failure modes and scoring rubrics.
The health push adds to OpenAI's wider medical efforts, including clinician-focused tools and separate research showing its o3 model helped identify 18 previously undiagnosed rare genetic diseases.
AI solved 5% of these rare disease cases. What will it take for this technology to help the other 95%?
When will insurance policies cover AI re-analysis for families still waiting on a rare disease diagnosis?
As AI solves medical mysteries, what ensures it doesn't create new biases or errors in patient care?
AI Accelerates Rare Disease Diagnosis: OpenAI’s o3 Model Cuts Diagnostic Turnaround by 25% at Boston Children’s Hospital
Overview
A major breakthrough in rare disease diagnosis has arrived, ushering in a new era for medical professionals. Recent research shows that commercial AI systems, building on earlier advances in large language models, now allow doctors to analyze patient genomes more quickly and accurately. This speeds up the diagnostic process, helping patients who often face long and difficult journeys to find answers. As a result, diagnostic delays are reduced and patient outcomes are improved, making advanced genetic analysis more accessible and effective for those with elusive conditions.