Microsoft Declares Independence: Three New In-House AI Models
Microsoft unveils MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — three foundational models built in-house.
The Announcement
Microsoft — the company that poured billions into OpenAI — just launched three fully in-house foundational models. Not wrappers. Not fine-tunes. Built from scratch.
Mustafa Suleyman called it the first salvo from Microsoft's superintelligence team and said renegotiating the OpenAI contract "unlocked Microsoft's ability to pursue superintelligence".
MAI-Transcribe-1
Lowest word error rate across 25 languages. Beats Whisper, GPT-Transcribe, and Gemini 3.1 Flash-Lite. Runs at 2.5x current Azure speed. Pricing: $0.36/hr.
MAI-Voice-1
Generates 60 seconds of audio in under one second. Custom voice cloning from a 10-second sample.
MAI-Image-2
Ranked #3 on Arena.ai. Rolling into Copilot Voice, Teams, Bing, and PowerPoint.