Microsoft says its MAI-DxO system diagnosed 83 percent of complex cases correctly, against 36 percent for primary care doctors and general internists, in a study published in Science, TechCrunch reports.
DxO coordinates several AI agents that act like a panel of doctors, questioning patients, ordering tests and debating diagnoses. The cases came from the New England Journal of Medicine, with emergency department cases added.
The doctors worked alone with reference materials. Specialists and physician teams, who normally handle such cases, were not in the comparison, and TechCrunch’s account does not give the sample size. That makes 36 percent a floor for a generalist working alone, not a measure of what a hospital team would score.
Cost is part of the pitch. With GPT-5, DxO cost about $3,000 less per case than GPT-5 alone, at roughly $8,400 a case, but raised accuracy only about four points. With GPT-5.5, accuracy rose from 72 percent to nearly 87 percent at a slightly lower cost.
Much of the gain comes from the underlying model.
Gains were larger for weaker ones, with Gemini 2.5 Flash rising from 46.4 percent to nearly 69 percent, and DxO added little on the emergency department cases.
Mustafa Suleyman, Microsoft’s AI chief, says he is chasing medical superintelligence. This was a simulation, not a clinical trial.
Real-world testing with Beth Israel Deaconess Medical Center is planned, and there are no results yet and TechCrunch gives no date. Until then, 83 percent describes a benchmark.
