Together AI launched a unified platform for building real-time voice agents by co-locating STT, LLM, and TTS infrastructure on a single cloud. This eliminates inter-vendor network hops and achieves end-to-end pipeline latency under 500ms. The platform now natively hosts Cartesia Sonic-3 (TTS) and Deepgram Nova-3/Flux/Aura-2 (STT/TTS) models alongside existing options like Whisper Large v3 and Kokoro. Key benefits include a single API, billing surface, and authentication layer, plus enterprise compliance features including SOC 2 Type II, HIPAA, zero data retention, and dedicated data residency. Decagon is cited as a production customer running enterprise customer support voice agents on the stack.
Table of contents
The architecture: Unified modular vs. multi-provider stacksModel choice on one platformVoice agent in production: DecagonGet started3 Impressions