Microsoft AI is deepening its ambitions in healthcare through a collaboration with Mayo Clinic to build a frontier AI model designed for clinical use. The companies said the effort will combine Mayo Clinic’s clinical expertise and de-identified health data with Microsoft’s AI infrastructure and model-building capabilities.
The announcement came alongside Microsoft AI’s release of seven new in-house models spanning reasoning, coding, image generation, transcription and voice. Together, the new models form what Microsoft calls its MAI model family, which the company says is intended to support a range of real-world tasks across business and consumer settings.
The healthcare model is intended to handle a broad scope of clinical reasoning and other medical workflows, according to Microsoft. The company said the system is expected to go beyond what current general-purpose AI tools can do in healthcare settings.
Microsoft said the model will first run inside Mayo Clinic’s own environment. There, it is expected to support use cases such as earlier diagnosis and more accurate treatment planning. If the model performs as expected, Microsoft said it could later be made available to other organizations through Microsoft Foundry.
The companies also said the model will be owned by Mayo Clinic, a structure they described as important for patient trust, clinical rigor, safety and responsible handling of health data.
The Mayo Clinic work sits within a broader strategy at Microsoft AI that emphasizes in-house development, enterprise data controls and models that can be customized for specific workflows. Microsoft said its new models are built on clean, traceable and licensed data, and are designed to work across Microsoft products and third-party developer platforms.
Among the newly announced systems is MAI-Thinking-1, which Microsoft describes as its flagship reasoning model. The company says it performs competitively within its size class on software engineering and math tasks. Other models include MAI-Code-1-Flash for coding tasks, MAI-Image-2.5 for image generation and editing, MAI-Transcribe-1.5 for transcription, and MAI-Voice-2 for speech generation.
Microsoft said some of these models will be available through Foundry as well as external platforms including OpenRouter, Fireworks and Baseten. The company also said developers will be able to tune the model weights themselves for the first time.
A major theme in Microsoft’s announcement is what it calls Frontier Tuning, a method for adapting models to real workflows using reinforcement learning in operational environments. The company argues that the most useful training data comes from the actual steps people take while completing work inside an organization.
Microsoft said this approach lets customers build models trained on their own data, within their own environments, while keeping that information under their control. The company cited internal and customer examples in which tuned models were more efficient and better aligned with enterprise requirements than general-purpose systems.
In the healthcare collaboration, Microsoft is extending that philosophy into a more sensitive domain. The company said some sectors, including health, require deeper collaboration because of the risks involved and the importance of accuracy, privacy and oversight.
Microsoft framed the new model releases and the Mayo Clinic partnership as part of a long-term plan to build systems that continuously improve as they receive more compute, better data and stronger evaluation methods.
The company said its goal is a future it calls Humanist Superintelligence, which it defines as advanced AI systems that remain tools under human control. In that vision, people retain authority over how the systems are used, especially in high-stakes areas such as medicine.
For now, the Mayo Clinic project represents one of Microsoft AI’s most visible steps into healthcare, pairing one of the world’s best-known hospital systems with a company pushing to expand its role in frontier AI development.