Mustafa Suleyman, CEO of Microsoft AI, told employees that an in-house model was better than a similar offering from OpenAI, one of the software giant’s largest customers and partners.
“We now have a pre-trained base model that is more powerful than the model that currently powers GPT 5.6 Sol,” Suleyman said during a town hall meeting with employees on Thursday. “And in fact, we used a quarter of the computation that OpenAI uses to train its model.”
Pre-training is the first step in building an AI model, when it learns patterns from huge amounts of text, images, and other information. This requires huge amounts of expensive computing power, so Microsoft could save money if it found a way to do it more efficiently.
Microsoft announced the model, MAI-Base-2, earlier Thursday.
Suleyman said Microsoft tested this model privately against OpenAI models, including one nicknamed “Spud” that powers GPT-5.6. He added that Microsoft’s model performed better in tests measuring how well AI can understand and predict information patterns.
The comparisons are based on technical information Microsoft has about the OpenAI models, Suleyman said.
“Privately and confidentially, we also compared it to OpenAI models,” Suleyman told employees, while suggesting that Microsoft has privileged access to information about the performance of OpenAI models.
It is notable that Suleyman chooses to highlight the limitations of OpenAI models, even in private discussions. Microsoft has invested billions of dollars in OpenAI and has a stake in the startup that is likely worth more than $100 billion. Any criticism of OpenAI could potentially reduce the value of this huge investment.
The two companies also have a major partnership, with Microsoft offering OpenAI models to its Azure cloud customers. Microsoft has also used OpenAI models for its own products such as Copilot.
But relations between the two companies have deteriorated over the past year. Increasingly, Microsoft is trying to distance itself from OpenAI, notably by working on its own models.
Suleyman also said that Microsoft’s new AI model used “12 times less calculations than what was used for DeepSeek, for example, on calculation efficiency.”
Suleyman did not specify which DeepSeek model, but a public blog post described MAI-Base-2 pre-training as much more efficient than a similar offering from the Chinese lab, called DeepSeek-V4-Pro-Base.
Microsoft in March created a new combined team for its commercial and consumer Copilot products and reassigned Suleyman to focus on building AI models in the company’s superintelligence unit.
Microsoft declined to comment and OpenAI did not respond to a request for comment Thursday.
Do you have any advice? Contact this reporter by email at astewart@businessinsider.com or Report to +1-425-344-8242. Use a personal email address and a non-work device; here is our guide to sharing information securely.
Gn bussni
