355
Audio & Video Production344
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics140
Design & Creative169
Customer Support131
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Operations & Admin87
Education & Learning131
At Build 2026, Microsoft announced MAI-Thinking-1 plus new models for images, speech, transcription, and coding, as it builds more AI in-house.
In short: Microsoft announced seven new in-house AI models at Build 2026, led by a new model called MAI-Thinking-1.
Microsoft used its Build 2026 developer conference to introduce a set of AI models it built itself. The main one is MAI-Thinking-1, which Microsoft calls its “flagship” model.
Microsoft says MAI-Thinking-1 is “medium-sized” and performs on par with other leading models on software engineering tests. A model is the engine behind an AI tool, like the “brain” that turns questions into answers. Microsoft also says it trained the model from scratch using “clean data,” and that it did not copy or “distill” (make a smaller version) from other companies’ models.
Alongside MAI-Thinking-1, Microsoft announced other models aimed at specific tasks. MAI-Image 2.5 and a “flash” version can create images from text prompts and edit images. MAI-Transcribe-1.5 is for turning speech into text, and Microsoft claims it is five times faster than competing models. MAI-Voice-2 and its flash version add 15 new languages and more voice options. Microsoft also announced MAI-Code-1, a coding-focused model that is designed to run efficiently and is integrated into GitHub Copilot and Visual Studio Code.
For everyday users, this could mean more AI features built into Microsoft products, and more choice in how those features work. It also signals Microsoft is relying less on OpenAI than it used to, after the companies recently renegotiated their partnership to loosen ties.
Source: The Verge AI