What is AssemblyAI
AssemblyAI sells pre-recorded, realtime, and synchronous speech-to-text APIs, plus speech-understanding and voice-agent capabilities. Self-serve access starts free and moves to usage-based billing without a contract.
Pre-recorded speech-to-text
As of October 17, 2026, Universal-3.5 Pro costs US$0.21 per audio hour and supports 18 languages, native code switching, and advanced speaker diarization. Universal-2 costs US$0.15 per hour and supports 99 languages.
For pre-recorded audio, speaker diarization costs US$0.02 per hour on either model. Keyterms Prompting and natural-language Prompting each cost US$0.05 per hour on Universal-3.5 Pro; Keyterms are included with Universal-2. Medical Mode costs US$0.15 per hour.
Realtime, sync, and voice-agent pricing
Universal-3.6 Pro Realtime, the current direct upgrade to 3.5 Pro, costs US$0.45 per hour and supports 32 languages. Universal-Streaming and Universal-Streaming Multilingual each cost US$0.15 per hour, while Whisper Streaming costs US$0.30 per hour. The Sync API costs US$0.45 per hour and processes clips up to two minutes in one request.
The managed Voice Agent API costs US$4.50 per connected hour, or US$0.075 per minute, billed by the second. It includes AssemblyAI's current realtime speech model, conversational Voice Agent LLM and low-latency Voice Agent TTS, plus prompting, voice focus, turn and interruption detection, recordings, transcripts, hosting, and session orchestration. The new Dictation API costs US$0.62 per hour for cleaned, instruction-shaped text from spoken utterances up to 120 seconds.
Custom enterprise arrangements provide tailored rate limits, concurrency, and deployment terms. Buyers should include any selected add-ons when estimating the effective audio-hour cost.
Official source: AssemblyAI pricing.