Microsoft Unveils MAI‑DxO: AI Diagnoses 4x More Accurately Than Doctors

Microsoft has introduced an experimental medical AI system called MAI‑DxO (Medical AI Diagnostic Orchestrator), which showed impressive diagnostic accuracy: 80–85% in complex cases, compared to just 20% by human doctors.

Unlike traditional tools, MAI‑DxO operates as a virtual medical board, combining the strengths of multiple large language models — GPT, Gemini, Claude, Llama, and Grok. Each model adds its perspective in a process called “chain-of-debate,” resulting in a refined final diagnosis.

In tests using 304 real clinical cases from the New England Journal of Medicine, MAI‑DxO significantly outperformed doctors. It was also 20% more cost-efficient in decision-making. Though still experimental and not yet approved by regulators, Microsoft plans further development and potential implementation in hospitals.