Microsoft Unveils MAI-DxO: AI Diagnostic Tool 3.7x More Accurate Than Doctors at Half the Cost
- Microsoft has developed an artificial intelligence diagnostic tool called MAI-DxO that outperformed human doctors by up to 3.7 times in identifying rare diseases and emergency medical conditions, according...
- The Microsoft AI (MAI) health team released SDBench and MAI-DxO on October 22, 2026.
- The tool achieved an 83 percent diagnostic accuracy rate in those trials.
Microsoft has developed an artificial intelligence diagnostic tool called MAI-DxO that outperformed human doctors by up to 3.7 times in identifying rare diseases and emergency medical conditions, according to reporting by yna.co.kr on October 23, 2026. Powered by an interactive sequential evaluation benchmark named SDBench, the system uses multiple AI panels to debate and verify diagnoses while cutting testing costs by half or more.
Microsoft AI Healthcare Team Unveils MAI-DxO and SDBench
The Microsoft AI (MAI) health team released SDBench and MAI-DxO on October 22, 2026. Unlike standard AI models that give immediate single-turn answers, MAI-DxO acts as a diagnostic orchestrator. Multiple AI agents function like a medical panel, proposing individual diagnoses, running tests, debating findings, and verifying results. The system also calculates the expected value and cost of medical tests to determine the most efficient diagnostic route.
Clinical Validation Results Against Human Doctors
The tool achieved an 83 percent diagnostic accuracy rate in those trials. That performance compares with 22 percent for general practitioners working without reference materials and 36 percent for doctors using textbooks and internet searches. In emergency room admission tests, MAI-DxO reached 80 percent accuracy compared to 47 percent for human physicians.
The AI system also lowered estimated diagnostic costs by half to two-thirds compared to standard hospital workflows by eliminating unnecessary tests. Mustafa Suleiman, CEO of MAI, described the achievement as an important step toward medical superintelligence
in comments cited by TechCrunch and reported by yna.co.kr.
Microsoft has partnered with Beth Israel Deaconess Medical Center in Boston to launch clinical studies on actual emergency room and hospital patients. The company has also made the SDBench evaluation code and cost-estimation algorithms publicly available to researchers, while consolidating its distributed health research teams into the MAI organization to speed up development. The complete study findings were published in the international scientific journal Science.
At the same time, tech publication reporting noted a methodological caveat in the trials. TechCrunch pointed out that the physicians serving as the control group in the clinical validation tests were internists and general practitioners, whereas complex diagnoses in real hospital settings are typically handled by specialists or multidisciplinary collaboration teams, according to yna.co.kr.