Capability change
- Before: Builders could not use these newly launched MAI audio models for real-time transcription or voice generation.
- Enabling change: Microsoft launched MAI-Transcribe-2-Streaming and MAI-Voice models; Vercel made Microsoft AI’s MAI audio models available through AI Gateway.
- After: Builders can access the MAI audio models, including through Vercel AI Gateway, for audio applications.
Human consequence: A builder can add these audio models to an application through a gateway rather than integrating only through a direct provider path.
Bottleneck: The items do not specify model access requirements, supported interfaces, or availability beyond the gateway announcement.
Buildability: B3. Verification: source-inspected. Related shift: compute-backed-ai-contracts.
The two updates connect a model launch with a serving route: Microsoft announced streaming transcription and voice models, while Vercel says MAI audio models are available through AI Gateway. That supports a concrete access change, not a claim about model quality or performance.
A builder could test the gateway route for transcription and voice features. Watch for details on availability, interfaces, and access requirements; the items do not establish those specifics.
Published automatically by the New Runtime weekly knowledge desk from the week's evidence.
Supporting evidence
- Microsoft launches MAI-Transcribe-2-Streaming Microsoft launched MAI-Transcribe-2-Streaming and two MAI-Voice models, offering real-time transcription, multilingual text-to-speech, low-latency audio generation, and voice cloning for conversational agents. (signal, 2026-10-03)
- We partnered with @MicrosoftAI to bring their secure, transparent, and cost-efficient models to AI Gateway. (signal, 2026-10-01)
Provenance
Auto-created by the New Runtime weekly knowledge desk on 2026-10-05 from evidence reviewed in the week of 2026-09-28. It is published through the live overlay and joins the hand-edited catalog at the next release.