Microsoft (MSFT.US) accelerates AI autonomy; self's self-developed image generation model to replace OpenAI-related技术

Mustafa Suleyman, Microsoft’s Head of AI, stated in an interview that compared to OpenAI’s models, Microsoft’s self-developed models run faster, cost less, generate higher-quality results, and also improve user retention rates.

In recent years, Microsoft has been developing large models to compete with OpenAI and Anthropic, aiming to gradually integrate AI capabilities into its product ecosystem and reduce the cost of AI services. Suleyman revealed that, for example, the operating cost of Microsoft’s self-developed Microsoft AI (MAI) model in PowerPoint has decreased by approximately 85% compared to before.

Although under the parties’ long-term cooperation agreement, Microsoft can currently use OpenAI’s large models for free, Microsoft needs to bear the significant computing infrastructure costs required to run these models. Therefore, adopting more efficient self-developed models is expected to bring considerable cost savings.

Suleyman pointed out that in office software such as PowerPoint, users often use AI to generate and edit images, such as creating infographics, and image generation is also one of the most commonly used AI features by new users. Therefore, improving image generation quality and response speed is crucial for enhancing user experience.

Besides image generation, Microsoft is also accelerating the rollout of other self-developed AI models. The company announced that enterprises including T-Mobile US (TMUS.US) and UK budget airline EasyJet have already started deploying Microsoft’s latest AI voice generation models in their customer service call centers.

Previous media reports indicated that Microsoft had started gradually replacing OpenAI’s and Anthropic’s text generation models in several Office software applications with its own self-developed models. Microsoft stated that its MAI model’s performance in common tasks in applications such as Excel has matched that of OpenAI’s latest GPT-5.6 model, but with higher cost-efficiency.

Suleyman said that the MAI model is now used in more than half of Microsoft’s products and is already being tested across all Microsoft products. He said: “This is a very strong start, and we will continue to push forward with full force.”