Microsoft Unveils 3 In-House AI Models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Microsoft launched three in-house AI models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—built by its superintelligence team, with MAI-Transcribe-1 ranked #1 on the FLEURS word error rate benchmark across 25 languages.
- Mustafa Suleyman said renegotiating Microsoft's contract with OpenAI "unlocked [Microsoft's] ability to pursue superintelligence" in a Verge interview, framing the launch as part of an "AI self-sufficiency" push.
- Microsoft currently "not able to build models in the very largest scale yet," but Suleyman said the "computation ramp is coming to enable us to do" it in 2026.
- Microsoft aims to create large cutting-edge AI models by 2027, per Bloomberg, and the new MAI models are now available to developers in Foundry.
- Suleyman had been preparing for his Microsoft AI role for 6-9 months before his job description was made public last month, per Hayden Field's reporting on The Verge interview.
Why it matters: Microsoft's three in-house models and Suleyman's 2027 frontier-model target show the company is operationalizing its split from OpenAI rather than just restructuring on paper. But Suleyman's own admission that Microsoft can't yet build at "the very largest scale"—with the gating compute ramp not arriving until 2026—means the independence story is conditional, not delivered.
Ask SkimNews
