Microsoft has reportedly begun deploying its own AI models across some of its most widely used productivity applications. Instead of relying mainly on models from OpenAI and Anthropic, the software giant is gradually replacing them with its in-house MAI (Microsoft AI) models for selected tasks, reports Bloomberg. While this still represents only a small share of Microsoft’s overall AI usage, it shows that the company’s proprietary models have moved beyond testing and are now handling real-world workloads.
At its Build 2026 developer conference, the company introduced seven MAI models, covering reasoning, coding, image generation, speech and transcription. The flagship MAI-Thinking-1 model is designed for complex reasoning tasks like coding, mathematics and multi-step problem-solving, while MAI-Code-1-Flash focuses on software development. The lineup also includes MAI-Image-2.5 for image creation and editing, MAI-Voice-2 for speech generation, and MAI-Transcribe-1.5 for speech-to-text applications.
Notably, cost is one of the biggest reasons behind Microsoft’s latest move. The company processes huge volumes of AI tokens across services like Copilot, Microsoft 365 and GitHub Copilot. Every AI response requires computing resources, making large-scale AI deployment expensive. Although Microsoft currently benefits from discounted access to OpenAI’s technology through its long-standing partnership, the company is preparing for a future where it may have to pay market rates for external AI models.
Earlier, Microsoft AI CEO Mustafa Suleyman has openly said the company wants to reduce, and eventually eliminate, the money it spends on Anthropic by replacing more workloads with MAI models. Developing its own AI also gives Microsoft greater control over pricing, performance, security and future product development.
However, the firm is not ending its partnerships with OpenAI or Anthropic, but it is changing how those models are used. Instead of depending on a single AI provider, the company is turning Copilot and Azure AI into multi-model platforms that can select the most suitable model for each task. Depending on the workload, customers may use Microsoft’s MAI models, OpenAI’s GPT models, Anthropic’s Claude models or other supported AI systems. This gives businesses greater flexibility to choose different models for coding, document analysis, reasoning or customer service without leaving Microsoft’s ecosystem.
The company has also started expanding MAI models across its broader product portfolio. They are already available in GitHub Copilot, where developers can choose Microsoft’s in-house models for AI-assisted coding. Additionally, Microsoft plans to introduce its own speech transcription model into Microsoft Teams and other products in the coming months. During Build, Microsoft also claimed that one of its coding models can match the programming capabilities of a previous-generation Anthropic Claude model while operating at a lower cost.
Microsoft’s latest AI push comes as competition in the industry intensifies. For example, Google continues to expand its Gemini models, Meta is investing heavily in the Llama family, Amazon is building its own AI ecosystem while partnering with Anthropic, and several Chinese AI companies are rapidly improving lower-cost foundation models.