Microsoft rolls out new AI models
Microsoft announced three new foundational AI models plus a 'mid-class' model as it pivots toward superintelligence—moves poised to reshape enterprise AI for climate analytics and accelerate demand for data‑centre capacity. The push follows a renegotiated OpenAI relationship and reflects a strategic bet on both high-end and mid-tier compute needs. ( )
MAI-Transcribe-1, MAI-Voice-1 and MAI-Image-2 were made generally available to developers through Microsoft Foundry and the MAI Playground on April 2, 2026. (microsoft.ai) Microsoft says MAI-Transcribe-1 supports the top 25 global languages, posts an average word‑error‑rate of 3.9% on the FLEURS benchmark and ranks first on FLEURS in 11 of those languages. (microsoft.ai) Microsoft reports MAI-Transcribe-1 delivers batch transcription speeds 2.5x faster than its previous Azure Fast offering and is priced at $0.36 per hour of audio in Foundry. (microsoft.ai) MAI-Voice-1 is billed as a “lightning‑fast” speech generator able to produce a full minute of audio in under one second on a single GPU, and Microsoft lists pricing starting at $22 per 1 million characters. (microsoft.ai) Microsoft says MAI-Image-2 is being phased into Copilot and Bing integrations and the company published token pricing for the model — roughly $5 per 1M input text tokens and $33 per 1M image-output tokens in Foundry. (siliconangle.com) Microsoft and OpenAI signed revised partnership terms in October 2025 that preserve Azure exclusivity around frontier models while also carving out rights that Microsoft says enable independent in‑house frontier work. (blogs.microsoft.com) Mustafa Suleyman, CEO of Microsoft AI, said the company is “not able to build models in the very largest scale yet” but that a compute ramp is coming in 2026–2027 and has warned frontier AI will require “hundreds of billions” of investment over the coming decade. (techmeme.com) (msn.com) Azure’s AI-driven growth has pushed Microsoft Cloud scale — Azure and related cloud revenue topped $50 billion in a quarter — while Microsoft has been increasing on‑premises GPU clusters (including Nvidia GB200 deployments) as it ramps capacity to support larger model training and inference. (geekwire.com) (theoutpost.ai) Microsoft’s broader AI efforts already target climate and geospatial use cases through projects such as the Aurora climate/forecasting initiative, the ClimaX foundation model for weather and climate, and a UNFCCC partnership to build a climate data hub, while its Climate Innovation Fund has deployed over $800m of a pledged $1bn into climate tech investments. (blogs.microsoft.com) (microsoft.github.io) (esgdive.com) (aimagazine.com)
