Hi Vapi team, I'd like to request that you add AssemblyAI's Universal-3 Pro Streaming model (

u3-rt-pro

) as a selectable option in the AssemblyAI transcriber config. Current limitation: The AssemblyAI transcriber in Vapi only exposes two models:

universal-streaming-english

and

universal-streaming-multilingual

. AssemblyAI launched Universal-3 Pro Streaming earlier this year (April 2026), and it's specifically designed for voice agents — better accuracy on names, account numbers, and short isolated turns. Concrete issue I'm hitting: I'm running a phone receptionist bot. When a caller responds to "Who do I have the pleasure of speaking with?" with a single-word name like "Devin," Universal-Streaming English transcribes it as the digit "7" (it hears "Devin" → "seven" → auto-formats to "7"). I confirmed in AssemblyAI's own playground that: - Universal-Streaming English fails on single-word "Devin" but works on "I'm Devin" - Universal-3 Pro Streaming transcribes single-word "Devin" correctly This is exactly the kind of voice-agent scenario u3-rt-pro was built for. Sample call ID where this happened:

019e0ac9-2fc2-7009-8eef-62d7f215cd0b

. The ask: Add

u3-rt-pro

to the allowed values for

transcriber.model

when

provider: assembly-ai

. Per AssemblyAI's docs, the API change should be straightforward — same WebSocket endpoint, just a different

speech_model

query param. I'd prefer to stay on AssemblyAI for the latency win (~300ms P50) but the older streaming model is causing real problems with name capture. Would love to see this added. Thanks!

Thank you for your detailed request and context. Currently, Vapi only supports

universal-streaming-english

and

universal-streaming-multilingual

as AssemblyAI transcriber models;

u3-rt-pro

is not yet available as a selectable option in the API configuration ([see documentation](https://docs.vapi.ai/api-reference/assistants/update-assistant)). Suggestion: Please submit this as a formal feature request via [Vapi Support](https://docs.vapi.ai/support) or the Vapi Discord—this helps prioritize new model integrations. We'll update the documentation and changelog when new models like

u3-rt-pro

become available. Source: - [Assistants API Reference – Update Assistant](https://docs.vapi.ai/api-reference/assistants/update-assistant)

@User can we get this transcriber model please?