A study published in Science reveals Microsoft’s AI diagnostic system correctly solves complex medical cases more than twice as often as generalist physicians, signaling progress toward what CEO Mustafa Suleyman calls 'medical superintelligence'.
- MAI-DxO achieved 83% diagnostic accuracy on complex cases vs. 36% for generalists
- System combines AI agents to simulate doctor panels and optimize cost and accuracy
- Microsoft aims to release a healthcare foundation AI model with Mayo Clinic in 2027
What happened
Microsoft announced results from a study published in Science showcasing its AI diagnostic system, MAI-DxO, outperforming generalist physicians on difficult medical cases. The system, which functions as a network of AI agents mimicking a panel of doctors, correctly diagnosed 83% of cases taken from the New England Journal of Medicine. This marks a significant improvement compared to the 36% accuracy achieved by primary care and general internal medicine doctors working individually with reference materials.
The evaluation involved multiple underlying AI models from companies including OpenAI, Google, Anthropic, and others. Microsoft’s best performance was achieved leveraging OpenAI’s GPT-5 and GPT-5.5 models. In addition to higher accuracy, the orchestrated system reduced diagnostic costs by thousands of dollars per case compared to AI models running alone. However, gains in emergency department scenarios were less pronounced.
Why it matters
This development addresses a pressing challenge in healthcare: the increasing gap between demand for medical expertise and the limited supply of specialists. By demonstrating that an integrated AI system can diagnose hard cases with greater accuracy and at lower costs than generalist doctors, Microsoft contributes to advancing medical AI tools that can augment clinical decision-making.
Microsoft’s work advances toward what CEO Mustafa Suleyman calls 'medical superintelligence'—a transformative AI capability that could fundamentally improve diagnostics and patient outcomes. The company’s collaboration with Mayo Clinic on a foundation healthcare model planned for release in 2027 aims to further enhance prediction accuracy and clinical foresight, potentially revolutionizing medical practice.
What to watch next
Key indicators to monitor include the launch and adoption of Microsoft’s healthcare foundation model expected next year, developed in partnership with Mayo Clinic. This model promises to push healthcare AI capabilities dramatically forward and form the infrastructure for breakthroughs in clinical decision support.
Additionally, the evolution of orchestrated AI diagnostic systems like MAI-DxO will be important to see, particularly how they integrate multiple AI models and maintain or improve cost-efficiency without sacrificing accuracy. Ongoing evaluation against more realistic clinical settings and specialist teams will also be crucial to validate practical benefits in healthcare delivery.