Microsoft’s In-House AI Models Now Rival OpenAI and Anthropic

Separate reports claim that Microsoft’s efforts to create in-house AI models that rival those from partner OpenAI have finally paid off. Its latest models are allegedly “competitive” with those offered by OpenAI and Anthropic.

“As we’ve said previously, we are using a mix of models, which includes continuing our deep partnership with OpenAI, along with models from Microsoft AI and open source models,” a Microsoft spokesperson said.

One report, from Bloomberg, is based on a single source, but it’s supported by a separate report in The Information that provides more details. According to that publication, Microsoft’s internal efforts to develop a new family of AI models is dubbed MAI (Microsoft Artificial Intelligence) and it’s part of Microsoft AI CEO Mustafa Suleyman’s plan to make the firm more self-sufficient. (We previously reported that Microsoft’s first model of this era was called MAI-1.)

Both publications note that Microsoft is working on reasoning models–the AI models that seem to take time to “think” and describe their process for getting to an answer–and that the software giant is testing how it might replace OpenAI’s models in its Copilot offerings. It is also apparently looking at models from other AI makers like DeepSeek, Meta, and xAI.

Microsoft’s partnership with OpenAI runs through at least 2030, but the two have already made a major change to the terms of their agreement. And Microsoft has been leery of the combustible company since it briefly ousted CEO Sam Altman in late 2023. Since then, it hired Mr. Suleyman, created Microsoft AI, and has been quite open about its need to diversify. And Microsoft CEO Satya Nadella last year referred to AI models as “a commodity,” pluggable and replaceable components that might be sourced from anywhere.

However things might change going forward, Microsoft’s internal AI model work has clearly reached an important internal milestone.