The Redmond, Washington-based tech giant, a major investor in OpenAI, has reportedly started testing models from xAI, Meta, and DeepSeek as potential alternatives to OpenAI’s technology in its Copilot platform. This move aligns with Microsoft’s broader strategy to reduce its reliance on OpenAI, despite the early partnership that positioned Microsoft as a leader in the competitive AI landscape.
In December, Reuters exclusively reported that Microsoft was exploring the integration of internal and third-party AI models into its flagship product, Microsoft 365 Copilot. This shift aims to diversify its AI infrastructure beyond OpenAI’s technology and lower operational costs. When Microsoft launched 365 Copilot in 2023, it heavily promoted the use of OpenAI’s GPT-4 model as a key feature.
According to The Information, Microsoft’s AI division, led by Mustafa Suleyman, has successfully trained a new family of models, internally dubbed MAI (Microsoft Artificial Intelligence). These models reportedly perform nearly as well as leading models from OpenAI and Anthropic on standard benchmarks. Additionally, the team is developing reasoning models that employ chain-of-thought techniques—a method that breaks down complex problems into intermediate reasoning steps—to directly compete with OpenAI’s offerings.
The report also notes that Suleyman’s team is experimenting with replacing OpenAI’s models in Copilot with the MAI models, which are significantly larger than Microsoft’s earlier Phi models. Microsoft is considering releasing the MAI models later this year as an application programming interface (API), enabling external developers to integrate these models into their own applications.