Microsoft launches its first advanced reasoning AI model

Microsoft has unveiled MAI-Thinking-1, its first in-house advanced reasoning AI model. Jay Peters reports for The Verge that the announcement came at Microsoft’s Build 2026 developer conference, alongside several other new models targeting image generation, transcription, voice, and coding.

Microsoft describes MAI-Thinking-1 as its new “flagship” model and a “medium-sized model” that “matches leading models” on “key” software engineering benchmarks. The company says it “trained it from the ground up on clean data, without distillation from third-party models.” That last point is notable: it means Microsoft did not use outputs from other AI systems, such as those from OpenAI, to train the new model.

The launch marks a significant shift for Microsoft. The company had long relied on models from OpenAI to power its AI products. It introduced its first in-house models only last year. Microsoft and OpenAI recently renegotiated their partnership agreement to reduce their mutual dependency.

A full suite of new models

MAI-Thinking-1 was not the only model Microsoft announced at Build 2026. The company presented a range of new tools covering several areas of generative AI:

• MAI-Image 2.5 (and a flash version): supports text-to-image generation and image editing.
• MAI-Transcribe-1.5: a speech-to-text model that Microsoft claims is “five times faster than competing models.”
• MAI-Voice-2 (and a flash version described as “coming soon”): adds 15 new languages and expanded voice options.
• MAI-Code-1-Flash: a coding model that Microsoft calls “inference-efficient,” meaning it produces results with relatively low computational cost. It is integrated into GitHub Copilot and Visual Studio Code.

The breadth of the announcements shows Microsoft’s ambition to develop its own AI capabilities across multiple product categories rather than depending on a single external partner.

For everyday users, the most visible impact will likely come through GitHub Copilot and Visual Studio Code, two widely used developer tools now powered by MAI-Code-1-Flash. The voice and transcription models could surface in consumer and enterprise products over time, though Microsoft has not announced specific integrations beyond the coding tools.

The reasoning capabilities of MAI-Thinking-1 place it in a competitive category that includes models from OpenAI, Google, and Anthropic. Reasoning models are designed to work through complex, multi-step problems more reliably than standard language models. Microsoft has not provided detailed independent benchmark results, so external verification of its performance claims is not yet available.

With this announcement, Microsoft positions itself as a full AI developer rather than primarily a distributor of other companies’ technology. Whether MAI-Thinking-1 can truly compete with the leading reasoning models on the market remains to be seen.