For your information, the MAI family arrived on June 2, 2026, when Msutafa Suleyman made the announcement about the seven new models for aspects like coding, reasoning, voice, transcription, image generation, and more. The flagship reasoning model dubbed MAI-Thinking-1 was trained from scratch on Microsoft’s clean and no-distillation data. Similarly. the MAI-Code-1-Flash runs 5 billion active parameters and is created specifically for GitHub Copilot and VS Code. As claimed by Microsoft, it is comparable to Haiku but cheaper.
