At the recently opened Build2026 Developer Conference, Microsoft unveiled a series of self-developed AI models. The launch of its first advanced reasoning model,
As a "medium-sized model" with 35 billion active parameters,
In addition to the major reasoning model, Microsoft also expanded the MAI family lineup for multimodal and vertical application scenarios. In the fields of image and voice, MAI-Image2.5, which supports text-to-image generation and image editing, along with its Flash version, was officially released; MAI-Transcribe-1.5, which features ultra-fast transcription capabilities, can reach five times the speed of competitors; and MAI-Voice-2, which now supports 15 additional languages (including an upcoming Flash version), significantly enriches the voice multimodal ecosystem.
For the developer ecosystem, the coding model MAI-Code-1, which features optimized reasoning efficiency, has been successfully integrated into GitHub Copilot and Visual Studio Code. This successful rollout of multiple self-developed models not only completes Microsoft's closed-loop ecosystem from underlying reasoning to upper-level applications but also highlights the strategic determination of a tech giant to pursue comprehensive AI self-reliance and reduce reliance on external technologies.
