Microsoft ships MAI models to rival Google, OpenAI
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Microsoft released three proprietary models — MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — publicly on Microsoft Foundry and the MAI Playground, with announcements posted by CEO Satya Nadella and the Azure account on April 2, 2026
- MAI-Transcribe-1 handles speech-to-text across 25 languages, runs 2.5 times faster than Microsoft's Azure Fast offering, and was built by a team of 10 people
- MAI-Voice-1 produces 60 seconds of natural-sounding audio in one second and supports custom voice creation from a short audio clip
- MAI-Image-2 has reached the top three on the Arena.ai image generation leaderboard, with rollouts already underway in Bing and PowerPoint
- All three models are priced below comparable offerings from Amazon and Google
- Microsoft was contractually barred from building frontier AI by a 2019 deal with OpenAI until October 2025; once that restriction lifted, it disclosed the MAI models had already been quietly powering Copilot and Teams behind the scenes
- Mustafa Suleyman, CEO of Microsoft AI, reaffirmed Microsoft's commitment to its OpenAI partnership even as the MAI family signals a parallel in-house strategy
Why it matters: Microsoft is now selling proprietary AI at prices below Amazon and Google while still depending on OpenAI for frontier models — a dual-track posture that became possible only after the October 2025 expiration of its 2019 non-compete with OpenAI. The same MAI stack already serving Copilot and Teams is now opened to outside developers on Foundry, giving Microsoft price leverage and an independent fallback against its most important AI partner.
Ask SkimNews
