Microsoft develops in-house AI models to compete with OpenAI 

Microsoft is developing in-house artificial intelligence (AI) reasoning models to compete with OpenAI, potentially reducing its dependence on the ChatGPT maker. The Redmond-based tech giant has been testing AI models from xAI, Meta, and DeepSeek as potential replacements for OpenAI’s technology in its Copilot products, according to the report. While Microsoft’s partnership with OpenAI helped position it as a leader in the AI space, the company has been working to diversify its AI model offerings.

In December, Reuters exclusively reported that Microsoft was adding internal and third-party AI models to power its Microsoft 365 Copilot, aiming to lower costs and reduce reliance on OpenAI’s GPT-4 model, which was a key selling point when Copilot was introduced in 2023.

According to The Information, Microsoft’s AI division, led by Mustafa Suleyman, has successfully trained a new family of models internally referred to as MAI. These models reportedly perform nearly as well as OpenAI’s and Anthropic’s leading models on common AI benchmarks. Suleyman’s team is also working on advanced reasoning models that utilise chain-of-thought techniques—an approach that improves intermediate reasoning capabilities when solving complex problems.

Microsoft is already experimenting with replacing OpenAI’s models in Copilot with its MAI models, which are significantly larger than its previous AI family, Phi. The company is also considering releasing MAI as an application programming interface (API) later this year, allowing developers to integrate the models into their own applications.

This strategic shift highlights Microsoft’s growing ambition in the AI industry, signaling a move toward greater independence while maintaining a competitive edge in the rapidly evolving AI market.