IT Home October 16 news: According to Business Insider, on the 15th local time, Microsoft AI business CEO Mustafa Suleyman said at an employee meeting that an AI model developed in-house by Microsoft performsbetter than OpenAI's comparable products, and its training cost is expected to be lower. "Our pre-trained foundational model outperforms the model currently powering GPT-5.6 Sol, and it actually usedonly a quarter of OpenAI's compute during training.。”
Pre-training is the first step in AI model development, during which the model learns patterns from massive amounts of text, images and other data. Since pre-training requires large amounts of compute and is costly, if Microsoft can improve training efficiency, it stands to substantially reduce costs.
Earlier that day, Microsoft released this in-house model, named MAI-Base-2.
Suleyman revealed that Microsoft had privately compared MAI-Base-2 with several OpenAI models, includingthe model codenamed Spud that underpins GPT-5.6. In tests of the AI's ability to understand and predict patterns in information, Microsoft's model achieved better results.
IT Home learned from the report that the comparison was based on technical documentation of OpenAI models in Microsoft's possession. Suleyman said Microsoft compared it with OpenAI's models "privately and under confidentiality," implying Microsoft possesses special information about OpenAI models' performance.
In addition, Suleyman also said Microsoft's new model has an advantagein compute efficiency as well: "Take DeepSeek as an example — we need only one-twelfth of the compute it requires." However, he did not reveal which DeepSeek model was the comparison target.