Microsoft's AI Diagnostic System Achieves 83% Accuracy in Complex Medical Simulations

Microsoft AI has released a new diagnostic system designed to rival the reasoning capabilities of experienced human doctors. The platform, known as MAI-DxO, was the subject of a study published this week in Science. The research indicates that the system correctly diagnosed complex medical scenarios significantly more often than generalist physicians. Mustafa Suleyman, the CEO of Microsoft AI, described the achievement as a pivotal step toward what the company is calling medical superintelligence.

The core of the system is a collaborative framework that coordinates multiple artificial intelligence agents. These agents function similarly to a panel of physicians who would question a patient, order necessary tests, and debate potential diagnoses before reaching a conclusion. In a simulation using difficult cases drawn from the New England Journal of Medicine, the system demonstrated a high level of precision. It arrived at the correct diagnosis in 83% of instances. By comparison, primary care doctors and general internists, working individually with standard reference materials, were correct in only 36% of those same cases.

A key caveat in the study is that such complex cases are typically handled by specialists or teams of physicians in real-world settings. Those specialists were not included in the comparison group. Dominic King, vice president of Health at Microsoft AI, noted that the results are substantial. He highlighted that a major challenge in healthcare is the growing gap between the demand for medical expertise and the available supply.

The data suggests that much of the performance improvement stems from the underlying power of the AI models themselves rather than the orchestration layer. Microsoft tested the system with models from several providers, including OpenAI, Google, Anthropic, Meta, and DeepSeek. The strongest results were achieved using OpenAI's GPT-5 and GPT-5.5 models. This finding places Microsoft in the same category as Google and OpenAI, which have also recently claimed that their AI tools can match or exceed human doctor performance.

Cost efficiency was another metric examined in the study. When using GPT-5 with a basic prompt, the system scored 78.9% on complex cases at an estimated cost of $8,400 per case. The MAI-DxO system reduced this cost by approximately $3,000 while improving accuracy by only about four points. However, for emergency department cases, there was little difference in either accuracy or cost between the standalone model and the orchestrated system.

The performance of the newer GPT-5.5 model showed a more distinct split in results. On complex cases, the diagnostic system raised accuracy from 72% to nearly 87% at a slightly lower cost. Yet, this gain did not appear in emergency department scenarios, where both configurations scored around 85%. The system proved most beneficial when paired with less capable models. For instance, Gemini 2.5 Flash saw its accuracy jump from 46.4% to nearly 69% with the assistance of the orchestration layer.

Microsoft researchers note that accuracy gains naturally diminish as models approach the theoretical maximum. A spokesperson explained that the orchestration helps optimize both accuracy and cost. Suleyman stated that 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 infrastructure will allow for enormous leaps in the field.

Suleyman, a former co-founder of DeepMind, expressed confidence in the rapid progress of healthcare AI. He expects models to become an order of magnitude better within the next two to three years. He cited the progress in coding and chatbots as evidence that such flywheels can accelerate quickly. The company is also partnering with Beth Israel Deaconess Medical Center in Boston to evaluate the system in real-world hospital settings, including emergency and urgent care environments.

This study coincides with a manifesto from Microsoft defining medical superintelligence as the ability to surpass the strongest human-led health systems in reasoning and coordination. The company aims to move beyond simple benchmarking and integrate deeply with clinical workflows. With existing tools for dictation and radiology, Microsoft sees this diagnostic system as a major expansion of its healthcare portfolio. Suleyman stated that the company is committed to having the leading clinical decision support products in the coming years.