Microsoft’s AI diagnostic system, MAI-DxO, correctly diagnosed complex medical cases more than twice as often as generalist physicians in a simulation, according to a study published in Science. Mustafa Suleyman, CEO of Microsoft AI, described the result as a step toward medical superintelligence.
DxO coordinates multiple AI agents that function like a panel of doctors, questioning patients, ordering tests and debating diagnoses. Using complex cases from the New England Journal of Medicine, it was correct 83% of the time, compared with 36% for primary care doctors and general internists working individually with reference materials. Specialists and physician teams, who typically handle such cases, were not included in the comparison.
The study’s data suggests much of the improvement comes from stronger underlying models, with DxO adding further gains. Microsoft tested models from OpenAI, Google, Anthropic, Meta and DeepSeek, achieving its best results with OpenAI’s GPT-5 and GPT-5.5. On complex cases, DxO lifted GPT-5.5 from 72% to nearly 87% at slightly lower cost, while GPT-5 gained about four points and roughly $3,000 in savings per case. Neither model saw meaningful gains on emergency department cases. DxO helped weaker models most, raising Gemini 2.5 Flash from 46.4% to nearly 69%. Researchers said accuracy gains naturally narrow as models near their theoretical limit.
Dominic King, vice president of Health at Microsoft AI, said the biggest challenge in healthcare is the growing gap between demand for and supply of medical expertise. Through a new partnership with Beth Israel Deaconess Medical Center in Boston, Microsoft will evaluate DxO in emergency and urgent care settings, comparing its recommendations against those of doctors treating consenting patients. King said he expects it to be the largest study of its kind. Google recently published a similar Beth Israel study of its AMIE diagnostic AI in The Lancet.
Suleyman said Microsoft is working with the Mayo Clinic to train a healthcare foundation model planned for release next year. He predicted a coding-style breakthrough for medical records in 2027 and expects healthcare models to become an order of magnitude more capable within 24 to 36 months, driven by a flywheel of real-world usage and iteration.
Suleyman, a DeepMind co-founder, and King previously launched DeepMind Health together. Microsoft already sells dictation and radiology tools to clinicians, and Suleyman said the company aims to lead in clinical decision support products over the next few years.
