Microsoft introduces MAI-Thinking-1 AI model at Build 2026 conference
What's the story
At the Build 2026 conference, Microsoft unveiled a range of new in-house artificial intelligence (AI) models. The highlight of the announcement is MAI-Thinking-1, a "flagship" model. This comes after the tech giant introduced its first in-house models last year, relying heavily on OpenAI's tools until now. The two companies recently renegotiated their partnership to loosen ties.
Model details
MAI-Thinking-1 is a 'medium-sized model'
Microsoft describes MAI-Thinking-1 as a "medium-sized model" that competes with top models on key software engineering benchmarks. The company claims it was "trained from the ground up on clean data, without distillation from third-party models." This unique approach sets MAI-Thinking-1 apart from other AI models in the market today.
Applications
Other models for image generation, transcription, and coding
Along with MAI-Thinking-1, Microsoft also unveiled several other AI models. These are designed for a variety of tasks such as image generation, transcription, voice, and coding. The MAI-Image 2.5 and its Flash version can perform text-to-image conversion and image editing. Meanwhile, the MAI-Transcribe-1.5 is touted to be "five times faster than competing models."
Voice and coding
MAI-Voice-2 adds 15 new languages
The new voice model, MAI-Voice-2, and its Flash version (coming soon) add 15 new languages and options for voices. The coding model, MAI-Code-1, is "inference-efficient" and integrated into GitHub Copilot as well as Visual Studio Code. These advancements showcase Microsoft's commitment to enhancing its AI capabilities across different domains.