Microsoft AI Beats Doctors on Hard Cases

Microsoft’s MAI-DxO artificial-intelligence diagnostic system correctly diagnosed complex medical cases more than twice as often as generalist physicians in a simulation, according to a study published Thursday in Science. The system reached 83% accuracy on complex cases drawn from the New England Journal of Medicine. Primary care doctors and general internists, working alone with reference materials, managed 36%. Cases of that complexity usually go to specialists or physician teams, but those teams weren’t included in the comparison. Mustafa Suleyman, Microsoft AI’s CEO, called the result a step toward “medical superintelligence.”

How the System Works

MAI-DxO, also called DxO, coordinates several AI agents that act like a panel of doctors. They question patients, order tests and debate diagnoses among themselves. Microsoft tested the system with models from OpenAI, Google, Anthropic, Meta and DeepSeek. Its strongest results came with OpenAI’s GPT-5 and GPT-5.5. The study data suggests stronger underlying models drove most of the improvement, while Microsoft’s orchestration layer added the rest. Microsoft joins Google and OpenAI in claiming its AI can match or outperform doctors. That’s a crowded claim to stake out.

Cost and Accuracy

Microsoft measured estimated operating costs alongside accuracy, and the numbers call for a closer look. With a short, basic prompt, GPT-5 scored 78.9% on complex cases at about $8,400 per case. DxO cut that cost by about $3,000 and raised accuracy by roughly four percentage points. On emergency department cases, GPT-5 with and without DxO showed little change in accuracy or cost. GPT-5.5, released in April 2026, scored 72% on complex cases on its own. With DxO, accuracy rose to nearly 87% at a slightly lower cost. Gemini 2.5 Flash scored 46.4% alone and nearly 69% with DxO. The gap between weaker and stronger models exceeded the boost DxO delivered, suggesting the foundation model still matters more than the wrapper. Microsoft researchers acknowledged that accuracy gains shrink as models approach the theoretical maximum. A company spokesperson said, “The DxO orchestration helps with the broader challenge of optimization both accuracy and cost.”

Commercial Ambitions and Real-World Testing

Dominic King, vice president of Health at Microsoft AI, told TechCrunch: “There is something very substantial and exciting here. The biggest challenge in healthcare delivery at the moment is the growing mismatch between demand and supply of medical expertise.” Suleyman said Microsoft is working with the Mayo Clinic to train a healthcare foundation model, with release planned for next year. He described DxO as part of the company’s foundational infrastructure for healthcare breakthroughs. “I really believe that in 2027, we will be on the verge of a coding style breakthrough for medical records,” Suleyman said. “I think we’re going to be able to stream highly accurate predictions about how a medical record might unfold, and obviously that kind of foresight for a clinician is gold dust.”

Microsoft plans to evaluate DxO’s clinical reasoning across hospital settings through a new partnership with Beth Israel Deaconess Medical Center in Boston. King said the technology is ready for real-world evaluation. Patients in the study will still receive standard care, including triage, specialist time, treatments and a decision about whether it’s safe to discharge them. Patients who consent will also engage with DxO, and researchers will compare its recommendations with those of human doctors. King said the study would include emergency and urgent-care settings and show whether the systems could be deployed at greater scale.

Competition and Strategy

Google published its own clinical study with Beth Israel in The Lancet this month, evaluating its AMIE conversational diagnostic AI during real-world urgent-care visits, with human physicians monitoring in real time. Microsoft’s Science study came alongside a manifesto describing medical superintelligence as “the capacity to surpass the strongest contemporary human-led health systems in medical reasoning and coordination across the full arc of care.” Microsoft says it wants to move beyond exam-style benchmarks toward integrating AI with clinicians and health systems. Suleyman said the company already sells dictation and radiology tools to clinicians and sees DxO as an extension of that business. “We have to have the absolute number one state-of-the-art medical clinical decision support products over the next few years,” he said.

Why This Matters: A system that gets 83% of hard cases right while costing around $8,400 per case raises questions about who pays and who bears the risk when it’s wrong. The savings Microsoft describes, about $3,000 per case in its own simulation, matter only if hospitals and insurers adopt the technology at scale. Those are private purchasing decisions, and the market will sort them out faster than any regulator could. But these results come from a simulation built on case studies, not patients in an emergency room. The Beth Israel trial will test the claim against real stakes. Until those results arrive, the claim that AI outperforms doctors remains the company’s assessment, and a careful buyer should treat it that way.