Microsoft Unveils Its First Advanced Reasoning AI

Microsoft unveiled a range of new in-house AI models at Build 2026, highlighting its latest flagship model: MAI-Thinking-1. This marks a significant advancement in Microsoft’s model development, following its initial in-house models introduced last year, after previously depending on OpenAI’s technology. The two companies have recently restructured their partnership to allow for more independence.

MAI-Thinking-1 is described as a “medium-sized model” that competes with top models on essential software engineering benchmarks. Microsoft claims it was developed entirely from scratch using clean data, without relying on third-party models.

In addition to MAI-Thinking-1, several other models were announced, focusing on various applications such as image generation, transcription, voice, and coding. MAI-Image 2.5 and its flash version offer capabilities for text-to-image generation and image editing. MAI-Transcribe-1.5 boasts a speed that is “five times faster than competing models.” The MAI-Voice-2 and its upcoming flash version will introduce 15 new languages and additional voice options. Lastly, the new coding model, MAI-Code-1-Flash, is designed for efficient inference and is integrated with GitHub Copilot and Visual Studio Code.