Microsoft has developed proprietary AI models, known as “MAI,” to compete with advanced systems from OpenAI and Anthropic.

These models have performed well in internal tests and are being evaluated for Microsoft’s Copilot AI assistants, which assist with tasks like document editing and scheduling.

Microsoft is also working on “reasoning models” to handle more complex queries, an area explored by OpenAI, Anthropic, and Alphabet.

The MAI initiative could reduce Microsoft’s reliance on OpenAI. Their US$13 billion partnership is being renegotiated, with OpenAI allowed to use other cloud providers until 2030.

Microsoft uses a mix of AI models, including in-house designs, OpenAI’s technology, and open-source options. It is also testing models from Meta, Anthropic, and xAI for potential integration into Copilot.

Microsoft’s AI portfolio includes smaller models called “Phi” and resold AI solutions. This investment in AI infrastructure aims to diversify its offerings and strengthen its position in the competitive market.

Recent Microsoft developments