Microsoft AI Tool Beats General Doctors in Medical Test Simulation

Microsoft says its new artificial intelligence diagnostic system correctly diagnosed complex medical cases in a simulation more than twice as often as generalist physicians. A study published in Science shows the system, named MAI-DxO or DxO for short, was correct 83 percent of the time in tests using complex cases from the New England Journal of Medicine. Primary care doctors and general internists working alone with reference materials were correct 36 percent of the time.

System Setup

The DxO system coordinates several AI agents that act like a team of doctors to question a patient, order tests, and debate a diagnosis. Mustafa Suleyman, CEO of Microsoft AI, called the results a step toward medical superintelligence. Dominic King, vice president of Health at Microsoft AI, said the biggest current challenge in healthcare is the mismatch between the demand and supply of medical expertise.

Model Performance

Study data indicates much of the improvement comes from underlying AI models rather than Microsoft's system alone. With a newer model named GPT-5.5 released in April 2026, DxO increased accuracy from 72 percent to nearly 87 percent at a lower cost, though emergency department case scores remained about 85 percent with or without DxO. Microsoft is working with the Mayo Clinic to train a healthcare foundation model for release next year, and has a new partnership with Beth Israel Deaconess Medical Center in Boston to test DxO in real hospital settings.

Context

AI diagnostic tools are being developed by several major tech companies to help address doctor shortages and heavy patient loads in hospitals. These systems use advanced language models to review patient symptoms and medical literature, though researchers note they still need extensive real-world testing before routine clinical use.