Microsoft says its AI diagnostic system correctly solved complex medical cases more than twice as often as general doctors in a simulation. The study was published in the journal Science, and Microsoft AI’s CEO Mustafa Suleyman called it a step toward “medical superintelligence”, TechCrunch reports. The headline number is striking, but the details matter.
How the system works
The system, called MAI-DxO, is made of several AI agents that act like a panel of doctors. They question a simulated patient, order tests and debate the diagnosis before settling on an answer.
On complex cases drawn from the New England Journal of Medicine, it was right 83% of the time. Primary care doctors and general internists working alone with reference materials were right 36% of the time.
The caveats
- The wrong comparison group, arguably. Cases this complex are usually handled by specialists or teams of doctors, and they were not part of the test.
- Most of the gain comes from the underlying model. Microsoft tested 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 with a basic prompt already scored 78.9% on complex cases. MAI-DxO added about four points and cut the estimated cost per case by about $3,000, from roughly $8,400.
- Little benefit on emergency cases. On emergency department cases, GPT-5.5 scored about 85% with or without Microsoft’s system.
- Bigger help for weaker models. Gemini 2.5 Flash went from 46.4% to nearly 69% with MAI-DxO, but even so, stronger models did better than any boost the system provided.
Next: real patients
Microsoft is partnering with Beth Israel Deaconess Medical Center in Boston to test the system in real hospital settings. Patients will get normal care from doctors, and those who agree will also interact with MAI-DxO so its recommendations can be compared with the doctors’. Google published a similar real-world study of its own diagnostic AI with the same hospital this month.
Suleyman also told TechCrunch that Microsoft is working with the Mayo Clinic on a healthcare AI model planned for release next year, and that he expects healthcare models to get “an order of magnitude better” within two to three years.
What it means for you
Nothing here is a tool you can use today, and an AI that scores well on test cases is not the same as one that is safe in a busy clinic. The realistic near-term future is AI as a second opinion that supports doctors on hard cases. If you use a chatbot for health questions, treat its answers as a starting point for a conversation with a professional, not a diagnosis.