Microsoft AI Claims Superior Efficiency Over OpenAI Models
CEO Mustafa Suleyman asserts MAI-Base-2 outperforms OpenAI counterparts using significantly less computing power.
1New Model Performance Claims
Microsoft AI CEO Mustafa Suleyman announced the development of MAI-Base-2 during a recent internal town hall meeting. He claimed this new pre-trained model demonstrates superior performance compared to the OpenAI model currently powering GPT-5.6. The evaluation focused on the model's ability to understand and predict complex patterns within large information datasets.
2Efficiency Gains in Training
A critical highlight of the announcement was the drastic reduction in resource consumption during the pre-training phase. Suleyman stated that Microsoft achieved these results using only one-quarter of the computing power required by OpenAI. This efficiency represents a significant financial optimization, as pre-training remains the most expensive stage of AI development.
3Internal Benchmarking Methodology
Microsoft conducted private tests comparing MAI-Base-2 against various OpenAI models, including one internally nicknamed Spud. These comparisons utilized technical data accessible to Microsoft through its partnership. Suleyman emphasized that the internal testing confirmed the new model's capability to outperform existing industry benchmarks in pattern recognition and information processing tasks.
4Comparison With DeepSeek Models
Beyond OpenAI, Microsoft also benchmarked its new model against external competitors. Suleyman noted that MAI-Base-2 utilized twelve times less computation than the DeepSeek-V4-Pro-Base model. This comparison underscores Microsoft's strategic focus on compute efficiency as a primary metric for evaluating the success of its superintelligence unit's latest technical developments.
5Strategic Partnership Dynamics
The public comparison is notable given Microsoft's multi-billion dollar investment in OpenAI and their deep integration. Microsoft relies on OpenAI models for Azure cloud services and its own Copilot products. However, the relationship has shown signs of fraying as Microsoft increasingly develops its own proprietary models to reduce dependency.
6Organizational Structural Shifts
In March, Microsoft consolidated its commercial and consumer Copilot teams to streamline operations. Mustafa Suleyman was reassigned to lead the superintelligence unit, specifically tasked with building internal AI models. This structural change reflects Microsoft's broader ambition to control its own foundational technology stack rather than relying exclusively on external partners.
7Financial and Market Implications
Microsoft holds a stake in OpenAI valued at over $100 billion, making any criticism a sensitive matter. Publicly questioning the performance of OpenAI models could potentially impact the perceived value of this investment. Despite this, Microsoft continues to prioritize internal innovation, signaling a shift toward a more independent AI development strategy.
8Future Industry Trajectory
The focus on training efficiency marks a pivotal shift in the AI race. As computing costs remain a barrier to entry, Microsoft's ability to produce stronger models with fewer resources provides a competitive edge. Future developments will likely center on whether these internal efficiency gains can be scaled across all commercial products.
Original article: Business Insider — Microsoft AI Claims Superior Efficiency Over OpenAI Models