IT Home, October 16: According to Business Insider, on the 15th local time, Mustafa Suleyman, CEO of Microsoft’s AI business, stated at an employee meeting that Microsoft’s internally developed AI model outperforms OpenAI’s equivalent product and has the potential for lower training costs.
“Our pre-trained foundational model outperforms the model currently powering GPT-5.6 Sol, and we used only one-fourth the computational power that OpenAI employed for training,” said Sulaiman.
Pre-training is the first step in AI model development, during which the model learns patterns from vast amounts of text, images, and other data. Since pre-training requires substantial computational power and is costly, if Microsoft can improve training efficiency, it could significantly reduce costs.
Earlier today, Microsoft released its proprietary model named MAI-Base-2.
Suleiman revealed that Microsoft privately compared MAI-Base-2 with several OpenAI models, including the model codenamed Spud, which powers GPT-5.6. During tests evaluating the AI's ability to understand and predict patterns in information, Microsoft's model achieved superior results.
IT Home learned from the report that the comparison was based on Microsoft’s access to OpenAI’s model technical documentation. Sulaiman stated that Microsoft had “privately and confidentially” compared its own models with OpenAI’s, suggesting that Microsoft possesses specialized information regarding OpenAI’s model performance.
In addition, Sulaiman stated that Microsoft’s new model also has advantages in computational efficiency. “For example, compared to DeepSeek, we require only one-twelfth of the computational resources.” However, he did not specify which DeepSeek model was used for the comparison.