TechCrunch reported on Thursday on a Science study of Microsoft's MAI-DxO, an AI diagnosis system. Several AI agents act as a medical panel. In a simulation, they quiz a patient, request tests, and argue toward a verdict. On hard New England Journal of Medicine cases it was right 83% of the time. Primary care doctors and general internists, working alone with reference materials, were right 36%. Microsoft AI CEO Mustafa Suleyman called it a step toward "medical superintelligence."
TechCrunch says specialists and physician teams, who usually handle such cases, were not compared. It says the study's data suggests much of the improvement comes from stronger underlying models, with DxO adding gains on top. The excerpt does not say which model produced the 83%. GPT-5 alone, with a short, basic prompt, scored 78.9% on complex cases at an estimated $8,400 per case to run. DxO cut about $3,000 and added about four points. GPT-5.5 rose from 72% to nearly 87%, and Gemini 2.5 Flash from 46.4% to nearly 69%. On emergency cases, gains were small.
The coordination layer helped most on complex cases and weaker models, and little on emergency department cases. Buyers can ask: Was it tested against specialists? Was it a simulation or real patients? Which model and prompt gave the headline score? What does the cost per case cover?