AI System Outpaces Physicians on Complex Diagnostics, Microsoft Claims

A new study published Thursday in Science reveals that a Microsoft AI diagnostic system, in a simulation, correctly identified complex medical conditions more than twice as often as generalist physicians. The system, called MAI-DxO, achieved an 83% accuracy rate, while primary care doctors and general internists, working individually with reference materials, were correct 36% of the time. Mustafa Suleyman, Microsoft AI’s CEO, described the result as a significant step toward creating “medical superintelligence.”

The MAI-DxO system operates by coordinating multiple AI agents that function as a virtual panel of doctors. These agents question a patient, order necessary tests, and debate the diagnosis to arrive at a conclusion. This multi-agent approach is designed to tackle difficult cases that are typically handled by specialists or physician teams, who were not part of the study's comparison group. The study's data indicates that while more powerful underlying AI models provide much of the improvement, Microsoft's orchestration system adds further gains on top.

Microsoft tested its system with models from several companies, but its best results were achieved using OpenAI's GPT-5 and GPT-5.5. The company also measured the cost of running the AI on each case. For instance, with GPT-5, the DxO system reduced the cost per case by about $3,000 while increasing accuracy by approximately four points. However, on emergency department cases, there was little improvement in either accuracy or cost compared to using GPT-5 alone. With the newer GPT-5.5 model, DxO increased accuracy on complex cases from 72% to nearly 87% at a slightly cheaper cost, but again, no significant gains were seen on emergency department cases.

The orchestration system appears to provide the most benefit for complex cases and when used with less capable AI models. For example, Gemini 2.5 Flash's score rose from 46.4% on its own to nearly 69% with DxO. Still, the performance gap between weaker and stronger models was larger than the boost provided by DxO. A Microsoft spokesperson stated, “The DxO orchestration helps with the broader challenge of optimization both accuracy and cost.”

In publishing these results, Microsoft joins Google and OpenAI in claiming that its AI can match or outperform doctors. Suleyman, who leads Microsoft's AI division, stated the company is working with the Mayo Clinic to train a foundation model for healthcare, with a release planned for next year. He believes this will enable the company to “make really enormous leaps here,” and said DxO is part of the foundational infrastructure Microsoft is building for healthcare breakthroughs.

Suleyman expressed confidence in the rapid advancement of this technology. “I really believe that in 2027, we will be on the verge of a coding style breakthrough for medical records,” he said. He added that he expects healthcare models to get “an order of magnitude better in capability within the next 24 to 36 months.” He noted that progress accelerates once a “flywheel” is established, where a model in production is used in the real world to identify and patch its own gaps.

To move toward real-world application, Microsoft is partnering with Beth Israel Deaconess Medical Center in Boston to evaluate the clinical reasoning capabilities of DxO across different hospital settings. Patients will receive standard care, while consenting participants will also engage with DxO. The recommendations from the AI will then be compared against those of human doctors. “This technology is ready for real-world evaluations,” said Dominic King, vice president of Health at Microsoft AI.

The study in Science is accompanied by a manifesto outlining what Microsoft believes is necessary to build medical superintelligence, defined as “the capacity to surpass the strongest contemporary human-led health systems in medical reasoning and coordination across the full arc of care.” The company's aim is to move beyond individual tasks and integrate with clinicians and healthcare systems to improve outcomes. Suleyman emphasized the importance of this goal, stating, “This is the kind of technology that everybody in the world wants, and that's why we're committed to doing everything we can to accelerate its introduction to the world.”

Microsoft already offers dictation and radiology tools to clinicians and sees DxO as an expansion of those efforts. Suleyman outlined a clear objective for the company's future in this area. “We have to have the absolute number one state-of-the-art medical clinical decision support products over the next few years,” he said.