Soniox · v6 · 3× · vu le 05 nov. 2026
52
Momentum
Soniox STT v6 is Soniox's current speech-to-text model, offered via API with real-time streaming (stt-rt-v6) and asynchronous file processing (stt-async-v6) across more than 60 languages. New features are agent text input and echo filtering, which removes a voice agent's own audio from recognition. No independent v6 benchmarks were published at the time of research.
Historique du momentum
07.08.05.11.
Fonctionnalités
| Real-Time Streaming | Yes: word-by-word streaming over WebSocket (stt-rt-v6), with mid-sentence finalization |
| Latency | No v6 figure published by vendor; v5 (Pipecat benchmark): 260 ms median TTFS, 305 ms P95 |
| License | Proprietary (hosted API) |
| Platform | Hosted API: WebSocket (stt-rt-v6) and REST (stt-async-v6); Python and Node SDKs |
| Price | $0.12/hour real-time streaming, $0.10/hour async (file); token-based |
| Languages | 60+ languages in one model, mid-sentence language mixing supported |