Microsoft is developing in-house artificial intelligence reasoning models to compete with OpenAI. The company may release these models as an application programming interface (API) to developers later this year, according to a report by The Information.
Based in Redmond, Washington, Microsoft has reportedly commenced testing models from xAI, Meta, and DeepSeek as potential replacements for OpenAI technology within its product Copilot. This move is part of Microsoft’s strategy to reduce its dependence on OpenAI while still leveraging AI technology.
In December, Reuters published that Microsoft has been working on integrating both internal and third-party AI models within its flagship product, Microsoft 365 Copilot. This shift aims to diversify its technology sources to lower costs and enhance performance.
When Microsoft 365 Copilot was introduced in 2023, one of its main features was that it utilized OpenAI’s GPT-4 model. Now, Microsoft’s AI division, led by Mustafa Suleyman, is reportedly training a new family of models called MAI.
These models are said to perform almost on par with leading alternatives from OpenAI and Anthropic based on widely accepted benchmarks. Microsoft’s team is also focusing on training reasoning models that employ chain-of-thought techniques, which generate intermediate reasoning to solve complex problems effectively.
The report indicates that Suleyman’s team is experimenting with replacing OpenAI’s models in Copilot with the larger MAI models. Furthermore, the company is considering making the MAI models available to external developers, allowing them to integrate these tools into their applications.
Microsoft and OpenAI have not commented yet on these developments.
