Microsoft Releases Three In-House MAI Foundational Models, Signaling Growing Independence from OpenAI

Microsoft's MAI Superintelligence team, led by CEO of Microsoft AI Mustafa Suleyman, released three foundational models: MAI-Transcribe-1 (speech recognition), MAI-Voice-1 (speech generation), and MAI-Image-2 (multimodal image generation), all available exclusively on Microsoft Foundry and the MAI Playground. Suleyman noted frontier-class language models are still 1–2 years away, but these releases give Microsoft its own multimodal capabilities independent of its $13B OpenAI partnership. The move is being closely watched by developers as a sign that the OpenAI-Microsoft exclusivity era may be ending, while Azure AI infrastructure costs continue to pressure margins.