An artificial intelligence model built by OpenAI has helped doctors put names to genetic diseases that had stumped them for years.
According to OpenAI, researchers used one of the company's "reasoning" models — a type of AI designed to work through problems in deliberate, step-by-step fashion rather than answering instantly — to assist physicians in diagnosing rare genetic conditions in children. The effort produced 18 new diagnoses in cases that had previously gone unsolved.
Rare genetic diseases are notoriously hard to pin down. Because each condition affects so few people, many doctors may never encounter a given disease in their careers, and families can spend years on what specialists call a "diagnostic odyssey," cycling through tests and appointments without an answer. A confirmed diagnosis matters: it can guide treatment, end uncertainty, and connect families to others facing the same condition.
OpenAI frames the work as a collaboration between AI and physicians rather than a replacement for them. In the described use, the model served as a tool to help doctors reach conclusions in cases that had resisted earlier attempts.
The announcement comes from OpenAI itself, so the results are presented through the company's own account rather than independent reporting or peer-reviewed publication noted in the source. The figure highlighted — 18 previously unsolved cases now diagnosed — is the central claim.
Why it matters: if AI reasoning models can reliably help untangle the hardest diagnostic cases, they could shorten the long, painful search that families of children with rare diseases endure today.