Baseten releases Voice Agent Stack and Realtime Voice API
Understand this story
Baseten has released Voice Agent Stack and Baseten Realtime Voice, both in private preview and requiring application via waitlist. The former is infrastructure for running production-grade voice agents on Baseten Dedicated Inference, and the latter is a voice agent API built on that stack. According to Baseten, Realtime Voice offers ultra-low turn latency, costs about $0.01 per minute of audio, and is compatible with the OpenAI Realtime protocol, so existing agents can migrate with only minor changes; calls handed off to larger inference models are billed per token, the same as standard LLM endpoints.
AI overview based on source reports · Updated 12 min ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Baseten BlogIntroducing the Baseten Voice Agent Stack: own and optimize voice agents end-to-end
Baseten launched the Baseten Voice Agent Stack, customizable low-latency infrastructure for running production voice agents on Baseten Dedicated Inference, alongside Baseten Realtime Voice, a reference agent built on it that uses an OpenAI Realtime-compatible API and costs about $0.01 per audio minute. Baseten reports Realtime Voice scores 93% on Full-Duplex-Bench and 73.5% on τ²-Bench Voice, with median turn latency staying near 1 second from 1 to 56 concurrent sessions. Both are in private preview with a waitlist.
Activity for this story
Not enough continuous observations to show a trend yet.