Microsoft AI Chief Claims In-House Model Beats OpenAI GPT-5.6 Using Quarter of Compute
Microsoft AI CEO Mustafa Suleyman told employees that the new MAI-Base-2 model outperforms the base model powering GPT-5.6 Sol while using only one-quarter of the compute. The claims are based on private tests and remain unverified by third parties.
ImportanceMaterialEvidenceE2 unreplicatedWrite-upQuick
Mustafa Suleyman, CEO of Microsoft AI, stated at an employee town hall on Thursday that the company's latest pre-trained base model, MAI-Base-2, is stronger than the model currently powering OpenAI's GPT-5.6 Sol. He claimed this performance advantage was achieved using only one-quarter of the computing power required to train OpenAI's model.
According to Business Insider, Suleyman revealed that Microsoft privately compared MAI-Base-2 against several OpenAI models, including one nicknamed "Spud" which underpins GPT-5.6. He asserted that based on technical information Microsoft holds about OpenAI's models, MAI-Base-2 performed better on tests measuring AI comprehension and pattern prediction. Additionally, he noted that the model used "12 times less computation" than DeepSeek-V4-Pro-Base regarding efficiency.
These comments highlight a shift in the relationship between Microsoft and OpenAI. While Microsoft remains a major investor and partner, holding a stake likely worth over $100 billion and integrating OpenAI models into Azure and Copilot, it has increasingly pursued independent model development. It is important to note that these comparisons rely entirely on Microsoft's internal testing and statements, with no independent verification or public benchmark data provided.