Microsoft AI Outperforms Generalists on Complex Medical Diagnoses

A new artificial intelligence system developed by Microsoft correctly diagnosed complex medical cases in a simulation more than twice as often as generalist physicians, marking a significant milestone in the race toward medical superintelligence. The findings were published this week in the journal Science, presenting data that suggests AI is increasingly capable of handling intricate diagnostic challenges that often stump human doctors working individually.

The system, known as MAI-DxO or DxO for short, functions by coordinating multiple AI agents that act similarly to a panel of specialists. These agents work together to question patients, order necessary tests, and debate potential diagnoses before arriving at a conclusion. In simulations utilizing complex cases drawn from the New England Journal of Medicine, the system achieved an accuracy rate of 83%. By comparison, primary care doctors and general internists, who worked individually with access to reference materials, were correct only 36% of the time. It is important to note that these specific cases are typically handled by specialists or teams in real-world settings, groups that were not part of this specific comparison.

Dominic King, vice president of Health at Microsoft AI, described the results as substantial and exciting. He highlighted that the primary challenge in current healthcare delivery is the widening gap between the demand for medical expertise and the available supply of clinicians. By publishing these results, Microsoft joins competitors like Google and OpenAI in asserting that their AI technologies can match or exceed human performance in specific diagnostic tasks.

The study’s data indicates that a significant portion of the performance improvement stems from the use of more powerful underlying AI models, with Microsoft’s orchestration layer providing additional gains. While the system was tested with models from various providers, including Anthropic, Meta, and DeepSeek, the highest accuracy scores were achieved using OpenAI’s GPT-5 and GPT-5.5 models. The research also examined cost efficiency, finding that while DxO could reduce costs on complex cases, the accuracy gains were sometimes modest when using the most advanced models.

For instance, using GPT-5 alone with a basic prompt yielded a 78.9% accuracy score at an estimated cost of $8,400 per case. The DxO system lowered that cost by roughly $3,000 but improved accuracy by only about four points. However, with the newer GPT-5.5 model, released in April 2026, the orchestration system boosted accuracy from 72% to nearly 87% at a slightly lower cost. The benefits were less pronounced for emergency department cases, where both standalone and orchestrated models performed similarly.

Mustafa Suleyman, CEO of Microsoft AI, stated that the company is collaborating with the Mayo Clinic to train a foundation model specifically for healthcare, with a release planned for next year. He believes this infrastructure will allow for major leaps in clinical decision support. Suleyman expressed confidence that by 2027, the technology will be on the verge of a breakthrough in processing medical records, enabling highly accurate predictions about how a patient’s condition might evolve.

To validate these capabilities in a real-world setting, Microsoft has partnered with Beth Israel Deaconess Medical Center in Boston. This new study will evaluate the clinical reasoning of DxO across various hospital environments, including emergency and urgent care. Patients will receive standard treatment while those who consent will also interact with the AI system. The results will be compared against the decisions made by human doctors. Microsoft views this as a critical step toward integrating AI into broader healthcare systems, aiming to surpass contemporary human-led health systems in reasoning and coordination.

Suleyman, who previously co-founded DeepMind and led Hark, emphasized that the company is committed to accelerating the introduction of this technology to the world. He noted that progress in AI often follows a rapid acceleration once real-world usage provides feedback for improvement. With existing tools for dictation and radiology, Microsoft sees DxO as a natural expansion, aiming to provide the most advanced clinical decision support products available in the coming years.