Microsoft AI, led by CEO Mustafa Suleyman, released three foundational models — MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — marking a significant step toward building proprietary AI capabilities beyond its $13 billion OpenAI partnership. The models, available through Microsoft Foundry, cover speech recognition, voice generation, and image creation at competitive prices.
MAI-Transcribe-1 delivers enterprise-grade speech-to-text across 25 languages at $0.36 per hour, running 2.5x faster than Microsoft's own Azure Fast offering and outperforming OpenAI's Whisper-large-v3 on 14 languages. MAI-Voice-1 generates 60 seconds of expressive audio in under one second and can clone a voice from just a few seconds of sample audio at $22 per million characters. MAI-Image-2 debuted as a top-3 model on the Arena.ai leaderboard with 2x faster generation times at $5 per million input tokens.
The launch signals Microsoft's growing independence from OpenAI as the AI race intensifies, with the company now powering its own Copilot, Bing, and PowerPoint products using these in-house models before making them available to external developers.
