Microsoft says an AI diagnostic system it built correctly solved complex medical cases in a simulation more than twice as often as generalist physicians. The findings appeared in a study published Thursday in Science. Mustafa Suleyman, chief executive of Microsoft AI, described the result as a step toward what he calls "medical superintelligence."
The system, named MAI-DxO, or DxO, coordinates several AI agents that function like a panel of doctors. They question a patient, order tests and debate a diagnosis. Using difficult cases drawn from the New England Journal of Medicine, it reached the right answer 83% of the time. Primary care doctors and general internists, who worked individually with reference materials, were right 36% of the time. Cases of this complexity are usually handled by specialists or teams of physicians, who were not part of the comparison.
"There is something very substantial and exciting here," said Dominic King, vice president of Health at Microsoft AI. He pointed to what he called the growing mismatch between demand and supply of medical expertise as healthcare delivery's biggest challenge. Microsoft joins Google and OpenAI in claiming its AI can match or beat doctors.
The study's own data suggests much of the gain comes from more capable underlying models, with Microsoft's orchestration layer adding improvements on top. Microsoft tested DxO with models from OpenAI, Google, Anthropic, Meta and DeepSeek, and got its best results with OpenAI's GPT-5 and GPT-5.5. GPT-5 alone, given a short, basic prompt, scored 78.9% on complex cases at roughly $8,400 per case. DxO trimmed about $3,000 from that cost but lifted accuracy by only around four points.
With GPT-5.5, a model released in April 2026, DxO raised accuracy on complex cases from 72% to nearly 87% at a slightly lower cost. On emergency department cases, however, neither model showed meaningful gains in accuracy or cost; GPT-5.5 scored about 85% with and without DxO. The system helped most with complex cases and weaker models: Gemini 2.5 Flash rose from 46.4% to nearly 69%. Even so, the gap between weaker and stronger models exceeded the boost DxO provided. Microsoft's researchers noted that accuracy gains naturally shrink as models near the theoretical maximum.
Suleyman said Microsoft is working with the Mayo Clinic to train a healthcare foundation model, planned for release next year. He expects 2027 to bring a coding-style breakthrough for medical records, with accurate predictions of how a record might unfold, which he called "gold dust" for clinicians. He also expects healthcare models to become "an order of magnitude better in capability within the next 24 to 36 months," driven by a flywheel of real-world usage and iteration.
Through a new partnership with Beth Israel Deaconess Medical Center in Boston, Microsoft will evaluate DxO's clinical reasoning across hospital settings. Patients will receive standard care, and those who consent will also interact with DxO, allowing its recommendations to be compared against human doctors'. King said the study would span emergency and urgent care settings and expects it to be the biggest of its kind. Google published a study this month in The Lancet, also with Beth Israel, evaluating its AMIE diagnostic AI in real-world urgent care visits under real-time physician monitoring.
The Science paper arrives alongside a Microsoft statement of what building medical superintelligence would require, defined as surpassing the strongest human-led health systems in medical reasoning and coordination across the full arc of care. Microsoft, which already sells dictation and radiology tools to clinicians, sees DxO as an extension of that work. "We have to have the absolute number one state-of-the-art medical clinical decision support products over the next few years," Suleyman said.
