Suleyman says Microsoft's MAI-Base-2 beats OpenAI's GPT 5.6 Sol model

Microsoft's relationship with OpenAI has been under strain for the past year or so, as Microsoft has increasingly tried to branch out and build its own models. In March, Microsoft 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.

Suleyman's comments to employees stand out because Microsoft has invested billions of dollars in OpenAI and owns a stake in the startup that is likely worth more than $100 billion, so any criticism of OpenAI could potentially dent the value of that investment. The two companies also have a major partnership, with Microsoft offering OpenAI models to its Azure cloud customers and using OpenAI models for products such as Copilot.

Suleyman also said Microsoft's new model used 12 times less computation than was used for DeepSeek, and a public blog post described the pre-training of MAI-Base-2 as more efficient than a similar offering from the Chinese lab, DeepSeek-V4-Pro-Base. Microsoft declined to comment and OpenAI did not respond to a request for comment on Thursday.