Microsoft is reportedly developing in-house artificial intelligence (AI) reasoning models to rival OpenAI, the company behind ChatGPT. The Information reported on Friday that Microsoft may offer these models to developers. This move signals Microsoft’s intent to lessen its reliance on the ChatGPT maker, even though their early partnership put Microsoft at the forefront of the AI race.
According to the report, Microsoft is already testing models from xAI, Meta, and DeepSeek as possible alternatives to OpenAI within its Copilot AI assistant. Reuters previously revealed in December that Microsoft has been integrating both internal and third-party AI models to power its flagship AI product, Microsoft 365 Copilot, to diversify its technology base and reduce costs. When Microsoft announced 365 Copilot in 2023, it highlighted its use of OpenAI’s GPT-4 model as a key selling point.
Microsoft is considering releasing the MAI models later this year as an application programming interface, which will allow outside developers to weave these models into their own apps, the report said.
The Information’s report states that Microsoft’s AI division, led by Mustafa Suleyman, has finalized the training of a family of models, internally called MAI. These models are reportedly performing almost as well as the top models from OpenAI and Anthropic on common benchmarks. The team is also working on reasoning models, which utilize “chain-of-thought” techniques that can generate answers through intermediate steps when addressing complex problems. The report suggests that these reasoning models could directly compete with OpenAI’s offerings.
Suleyman’s team is already experimenting with substituting the MAI models, which are larger than Microsoft’s earlier Phi models, for OpenAI’s models in Copilot. The report also indicates that Microsoft is contemplating releasing the MAI models later this year as an application programming interface (API), allowing external developers to incorporate these models into their own applications.
Microsoft and OpenAI have not yet responded to requests for comment from Reuters.
