Deepgram
#6 in Transcription (STT)deepgram · since 2025-02-12 · 11× · last seen Jun 30, 2026
22
Momentum
Deepgram is a commercial Voice AI platform whose core offering is a Speech-to-Text (STT) API. The current flagship model Nova-3 was launched in February 2025 and supports 45+ languages for both batch and real-time streaming transcription. Additional offerings include the "Flux" conversational STT model (with turn detection), Text-to-Speech (Aura-2), and a Voice Agent API. Deployment is available as a cloud API or self-hosted (on-premises). Voice cloning is not an offered feature.
Momentum trend
04.04.03.07.
Features
| Latency | Transcript in under 300 ms (P50 median per vendor). Deepgram states <300 ms as target for streaming STT. |
| License | Proprietary/commercial (SaaS). Self-hosted option for enterprise requires separate API key and distribution credentials. |
| Platform | Cloud API (REST + WebSocket) as well as self-hosted (on-premises via container images). SDKs for JavaScript, Python, .NET, Go, Java. |
| Release Date | Nova-3 (current flagship model): February 12, 2025 |