Details
- Mustafa Suleyman says Microsoft’s first reasoning model, MAI-Thinking-1, is built from scratch and is now available in Microsoft Foundry.
- Microsoft’s official MAI announcement says the model is in public preview on Microsoft Foundry, with a try-it-now path also surfaced in Microsoft AI materials.
- Microsoft describes MAI-Thinking-1 as its first dedicated reasoning model, aimed at stronger math, coding, and general reasoning tasks.
- The model card and related Foundry pages frame it as a medium-sized sparse mixture-of-experts model with about 35 billion active parameters and roughly 1 trillion total parameters.
- Microsoft positions the model for enterprise deployment through Foundry, signaling closer productization for developers and customers already using Azure and Microsoft AI tools.
- The post’s “Try it here” link points users toward direct experimentation, suggesting Microsoft is prioritizing immediate adoption rather than a long rollout.
Impact
The release strengthens Microsoft’s effort to build a more independent model stack inside Foundry, reducing reliance on external frontier providers for reasoning workloads. It also raises competitive pressure on enterprise AI platforms by pairing a proprietary reasoning model with Microsoft’s distribution and cloud tooling. In a market where pricing, latency, and deployment control matter, a first-party model like MAI-Thinking-1 helps Microsoft compete on both product integration and cost structure.
