Microsoft’s Mustafa Suleyman says its AI outperforms doctors on hard cases as he chases ‘medical superintelligence’

Microsoft’s MAI-DxO diagnostic system correctly identified 83% of simulated complex medical cases drawn from the New England Journal of Medicine, compared with 36% for individual primary care doctors and general internists using reference materials, according to a study published in Science. MAI-DxO coordinates multiple AI agents that question patients, order tests, and debate diagnoses. The physician comparison did not include specialists or physician teams, which commonly handle cases of this complexity. Microsoft tested the system with models from OpenAI, Google, Anthropic, Meta, and DeepSeek. GPT-5 alone scored 78.9% on complex cases at an estimated $8,400 per case; MAI-DxO reduced estimated cost by about $3,000 and increased accuracy by roughly four percentage points. With GPT-5.5, MAI-DxO raised complex-case accuracy from 72% to nearly 87% at slightly lower cost, while it produced little accuracy or cost improvement on emergency-department cases. Microsoft AI CEO Mustafa Suleyman called the work a step toward “medical superintelligence.” Microsoft plans to release a healthcare foundation model trained with Mayo Clinic next year and will evaluate MAI-DxO with Beth Israel Deaconess Medical Center in hospital and urgent-care settings. Patients will receive standard care, while consenting participants also interact with the system; researchers will compare its recommendations with clinicians’ decisions.