AI Diagnostics: Microsoft System Beats Doctors in Complex Cases

An artificial intelligence diagnostic system achieved an 83% accuracy rate on complex medical cases in a simulation, more than double the 36% success rate of generalist physicians. This striking performance gap was highlighted in a study published Thursday in Science, which positions the technology as a potential shift in how healthcare providers handle difficult diagnoses. Mustafa Suleyman, the CEO of Microsoft AI, described the findings as a significant step toward achieving what he terms medical superintelligence.

The system, known as MAI-DxO, functions by coordinating multiple AI agents that mimic a panel of doctors. These agents engage in questioning patients, ordering tests, and debating potential diagnoses. The simulation utilized complex cases from the New England Journal of Medicine. While the results show a clear advantage for the AI over primary care doctors and general internists working individually, the study notes that such complex cases are typically handled by specialists or teams in real-world settings, who were not included in this specific comparison.

Microsoft joins Google and OpenAI in asserting that its AI can match or exceed human medical expertise. However, the data suggests that much of the improvement stems from powerful underlying models rather than the orchestration system alone. Microsoft tested the framework with models from OpenAI, Google, Anthropic, Meta, and DeepSeek. The best results were seen with OpenAI’s GPT-5 and GPT-5.5, though the company noted that the orchestration layer added incremental gains on top of the base model performance.

Cost and accuracy were both measured in the study. GPT-5 alone scored 78.9% on complex cases at an estimated cost of $8,400 per case. The DxO system reduced this cost by approximately $3,000 while raising accuracy by only four points. For emergency department cases, the improvements in both cost and accuracy were minimal. With the newer GPT-5.5 model, released in April 2026, the system improved accuracy from 72% to nearly 87% on complex cases at a slightly lower cost, but again saw little change in emergency scenarios.

Dominic King, vice president of Health at Microsoft AI, stated that the technology addresses the growing mismatch between the demand for medical expertise and its supply. The company is currently working with the Mayo Clinic to train a foundation model for healthcare, with a release planned for next year. Suleyman expressed confidence that this model will enable substantial leaps in the field, predicting a coding style breakthrough for medical records by 2027. He expects healthcare models to become an order of magnitude better in capability within the next 24 to 36 months.

To validate these findings, Microsoft is partnering with Beth Israel Deaconess Medical Center in Boston to evaluate the clinical reasoning capabilities of the system across different hospital settings. Patients will receive standard treatment while those who consent will also engage with the AI. The goal is to compare the AI’s recommendations against human doctors’ decisions in real-world urgent care and emergency environments. King described this as a major study that will provide compelling evidence for the future deployment of such systems at scale.

Microsoft views this work as an expansion of its existing sales of dictation and radiology tools to clinicians. Suleyman believes the focus on medical superintelligence could differentiate Microsoft from other frontier labs. He emphasized that the company is committed to building the number one state-of-the-art medical clinical decision support product in the coming years, aiming to integrate these systems deeply into healthcare workflows for better patient outcomes.