Microsoft AI models are now available on AI Gateway
Microsoft AI models are now accessible via Vercel’s AI Gateway, offering secure, zero-data-retention speech generation and live transcription with multilingual support and low latency.
Vercel and Microsoft AI have integrated MAI models into AI Gateway, a platform designed to provide controlled access to AI services. This partnership emphasizes security and zero data retention, aligning with Vercel’s commitment to data transparency and user control. AI Gateway now supports MAI’s latest audio models, expanding its capabilities for developers seeking reliable AI-powered speech solutions.
Microsoft’s MAI-Voice-2.1 and MAI-Voice-2.1-Flash enable speech generation in 23 languages, with the former maintaining consistent speaker identity for longer passages. These models cater to applications like narration, audiobooks, and podcasts where expressive delivery is critical. The Flash variant prioritizes speed, making it suitable for voice agents and assistants requiring low-latency responses.
MAI-Transcribe-2 Streaming delivers real-time transcription by processing audio as it arrives, allowing live captions or conversation tracking without waiting for completion. This feature is particularly useful for applications needing immediate text output, such as live events or customer service interactions. The model returns partial transcripts that may update as more audio is received.
AI Gateway bills users at the listed model rates without additional platform fees or markup on inference costs. Developers can integrate these models using AI SDK 7, with specific endpoints for speech generation and transcription. The platform also offers unified API access, usage tracking, cost monitoring, and failover support across providers, simplifying deployment and management.