Build 2026: Microsoft Launches First Flagship Reasoning AI Model and More

Microsoft announced today the release of new in-house MAI models from its AI Superintelligence team. The main announcement is MAI-Thinking-1, which is the company’s first reasoning model designed to offer high efficiency at a low-token cost.

MAI-Thinking-1 is available in private preview via the Microsoft Foundry platform, and Microsoft said that it matches Anthropic’s Claude Opus 4.6 model released in February on coding abilities on the SWE Bench Pro benchmark.

The company also announced the following updates for its MAI models:

MAI-Image-2.5 (+ Flash variant): This new text-to-image and image-to-model is now available in PowerPoint, Microsoft Foundry, and it’s coming soon to OneDrive. According to Daigle, it surpasses the quality of Google’s Nano Banana Pro model released in November on the ELO benchmark.

MAI-Transcribe-1.5: This new transcription model offers “state-of-the-art accuracy” across 43 languages.

MAI-Voice-2 (+ Flash variant): These two natural speech generation models are now available in more than 15 additional languages with new voice options.

MAI-Code-1: This inference coding model is now available in GitHub Copilot and VS Code.

Lastly, Microsoft said that its in-house AI models will soon become available on more platforms, including Fireworks AI, Baseten, and OpenRouter. “Fireworks AI is now generally available on Foundry, giving developers a single platform experience with enterprise governance and Azure data residency, regardless of the model they choose,” GitHub COO Kyle Daigle also announced today.