Deepgram vs AssemblyAI — rates and how they count
Prices checked on each vendor's pricing page in November 2026.
AI Tools Hub Editorial · published · updated
Rates
| Deepgram | AssemblyAI | |
|---|---|---|
| Unit | Minute | Hour |
| Pre-recorded, monolingual | Nova-3 $0.0043/min ($0.258/hr) | Universal-2 $0.15/hr |
| Pre-recorded, multilingual | $0.0052/min ($0.312/hr) | Universal-3.5 Pro $0.21/hr |
| Streaming | $0.0077–$0.0092/min | $0.15–$0.45/hr, billed on connection time |
| Free | $200 credit, no expiry | $50 credit |
| Self-hosted | Enterprise | Not stated |
| Extras | TTS, voice agents | PII redaction, summarisation, LLM gateway |
Streaming is billed differently
AssemblyAI bills streaming on how long the WebSocket stays open, not how much audio arrives. A voice agent holding a connection through long silences pays for the silence. Model this against real session lengths before comparing headline rates.
Redaction before storage
AssemblyAI includes PII redaction at transcription time. Redacting afterwards means the unredacted text existed somewhere, which is a different compliance position.
Test on your own audio
Published accuracy figures are measured under conditions each vendor chose and are not comparable. Both offer free credits; run the same file through each, mark errors in the first five minutes, and count only the errors that change meaning.
FAQ
Which handles Japanese better?
There is no independent comparison. Use the free credits on your own audio.