Microsoft announces seven AI models, including "MAI-Thinking-1," which matches Claude Sonnet 4.6 in performance, and the voice cloning model "MAI-Voice-2"

Microsoft unveiled seven in-house AI models spanning reasoning, coding, image generation, transcription, and speech synthesis.

The lineup includes MAI-Thinking-1, MAI-Code-1-Flash, MAI-Image-2.5 and Flash, MAI Transcribe-1.5, and MAI-Voice-2 and Flash.

Microsoft says MAI-Thinking-1 beat Claude Sonnet 4.6 in human evaluation, while the speech and vision models target multilingual and faster-use cases.

Several models are already available or will roll out through Microsoft Foundry, Visual Studio Code, and GitHub Copilot.