OpenAI’s o3 Model Helps Identify 18 Rare Disease Cases in Children
Researchers at Boston Children’s Hospital and Harvard have used OpenAI’s o3 Deep Research model to uncover 18 new diagnoses among children with rare genetic diseases that had remained unexplained for years. The findings appeared in a study published Wednesday in NEJM AI.
AI Helps Solve Longstanding Medical Mysteries
The research team revisited 376 de-identified pediatric cases that had already undergone extensive genetic testing and expert evaluation without reaching a diagnosis. By combining genetic information, clinical symptoms, and medical literature, the o3 Deep Research model helped clinicians identify conditions across several categories. These included neurodevelopmental disorders, rare neuromuscular diseases, sudden unexpected death in pediatrics, and early-onset psychosis.
Rather than acting as a consumer diagnostic tool, the system supports clinicians and researchers as they navigate complex medical data. As a result, doctors can analyze difficult cases more efficiently and uncover connections that may otherwise remain hidden.
“We combine genetic information, phenotypic information, literature search, and the reasoning of AI to deliver diagnoses to families that were once left without any answers,” Brownstein said.
Expanding the Role of AI in Healthcare
The peer-reviewed study strengthens a broader AI initiative underway at Boston Children’s Hospital. Earlier this year, the hospital reported that its AI programs had already contributed to more than 40 rare disease diagnoses previously considered unsolvable. Meanwhile, its collaboration with OpenAI has continued since early 2025, supported by a $50 million commitment.
Additionally, the hospital has integrated AI tools across many departments. More than one-third of employees now use AI in their daily work. Consequently, the organization estimates it has saved around 60,000 hours through AI-assisted workflows, allowing resources to be redirected toward other priorities.
New Hope for Rare Disease Families
Rare diseases affect an estimated 300 million people worldwide. However, many conditions occur so infrequently that even specialists struggle to identify them. Therefore, families often face years of uncertainty before receiving an accurate diagnosis.
“This was unthinkable before, but is now providing hope to so many families,” Brownstein said.
Researchers stressed that AI serves as a tool to support medical experts rather than replace them. Every diagnosis still required clinician review, verification, and interpretation before reaching patients and their families.
