How to Install / Upgrade: Microsoft launches new in-house AI models it says cut costs up to 89% versus OpenAI

Microsoft AI put two new in-house models into public preview: MAI-Image-2.5-Pro, its highest-fidelity image generator to date, and MAI-Voice-2-Flash, a speech model built for high-volume enterprise workloads. Microsoft says the models cut costs up to 89% versus OpenAI. Both are available now in public preview rather than as a finished general release.

To try them, request or enable access to the public preview for Microsoft’s in-house AI models, then point image-generation work at MAI-Image-2.5-Pro and high-volume speech work at MAI-Voice-2-Flash. Keep existing OpenAI or other model paths in place until you have confirmed the preview works for your use cases. Switch traffic only after you have validated quality and cost on a limited workload.

Public preview means features, limits, and pricing can still change, so do not treat either model as production-final without your own checks. Confirm you are calling MAI-Image-2.5-Pro for images and MAI-Voice-2-Flash for speech, and measure output quality and cost against your current OpenAI baseline before a full cutover. Keep a rollback path ready if preview access, quality, or cost does not match what you need.