Microsoft has announced seven new Microsoft MAI models at its Build 2026 developer conference in San Francisco. These in-house systems span image, voice, and transcription capabilities to support developer workflows. Consequently, the release marks a significant expansion of the company’s proprietary technology portfolio. The models were developed by the Microsoft AI Superintelligence team, which was formed in November 2025.
The New Microsoft MAI Models
Specifically, the lineup includes MAI-Thinking-1, which is the first reasoning model developed by the company. This 35 billion active parameter model handles complex multi-step instructions and long-context reasoning. It was built from the ground up on clean data, without distillation from third-party frontier models. Meanwhile, MAI-Code-1-Flash provides fast assistance for developers within GitHub Copilot in Visual Studio Code.
Reasoning and Coding Capabilities
In addition, the company introduced MAI-Image-2.5 and its flash variant for text-to-image and image-to-image workloads. These tools are already live in PowerPoint and rolling out on OpenDrive. Furthermore, MAI-Voice-2 adds support for more than 15 languages, including French, Spanish, German, and Hindi. This voice model features a wider emotional range that covers tones such as angry, confused, and embarrassed.
Image and Voice Enhancements
Notably, the new MAI-Transcribe-1.5 model trims latency to under 200 milliseconds for most languages. It combines accuracy across 43 languages, with streaming capabilities planned for a future release. According to the company, these Microsoft MAI models demonstrate high efficiency compared to existing market alternatives. For instance, MAI-Code-1-Flash outperforms Claude Haiku 4.5 across all core coding benchmarks tested.
Performance Benchmarks and Transcription
“This is about solving real concrete problems and doing it in such a way that it remains grounded and controllable.”
Mustafa Suleyman, CEO of Microsoft AI
Strategic Shift in AI Development
Notably, the development of these systems represents a strategic shift in the company’s AI strategy. By building proprietary foundational systems, the technology giant is reducing its reliance on external partnerships. As a result, this move positions the company as a direct competitor to other major AI developers. This transition follows a period of high capital expenditures, which reached 37.5 billion dollars in a single quarter.
Currently, the image, transcription, and voice systems are available on Microsoft Foundry and the MAI Playground. The company plans to expand availability to external platforms like Fireworks AI, Baseten, and Open Router. Therefore, developers can begin integrating these Microsoft MAI models into their applications immediately.
Source: X (@holgermu) — researched
