Cozzo

Best Speech-to-Text APIs for Voice Agents (2026)

Voice agents need real-time transcription that knows when the caller has finished speaking. Prices below are each vendor's published pay-as-you-go rates. Units differ (per minute, per hour, per token), and streaming is often billed on connection time, not speech.

Deepgram, AssemblyAI, Speechmatics, Gladia and Soniox compared on published real-time transcription pricing.

ToolStarting priceBest for
Deepgram logoDeepgramNova-3 streaming $0.0048/min ($0.29/hr); $200 free creditLow-latency speech-to-text for voice agents
AssemblyAI logoAssemblyAIStreaming from $0.15/hr; $50 free creditsAccurate, inexpensive transcription
Speechmatics logoSpeechmaticsFrom $0.24/hr pay as you go (as low as ~$0.12/hr with discounts)Multilingual transcription with language switching
Soniox logoSoniox~$0.10/hr async; ~$0.12/hr real-timeVery low-cost speech-to-text
Gladia logoGladiaAsync from €0.16/hr; real-time €0.68/hrReal-time multilingual transcription

STT models compared

ModelPriceLatencyLanguages
Deepgram Flux$0.0065/min (English)Built-in turn detectionEnglish or multilingualSource
ElevenLabs Scribe v2 Realtime$0.39 / hr~150 ms partials90+Source
Deepgram Nova-3$0.0048/min streaming (monolingual, promo; reg. $0.0077)—45+ (Nova models)Source
AssemblyAI Universal-3.5 Pro$0.21 / hrAsync (pre-recorded)18Source
OpenAI gpt-realtime-whisper$0.017/minLive transcription—Source
Speechmatics Linden 1$0.30 / hr real-timeBuilt-in turn detection and VAD (real-time)55+Source
Speechmatics Melia 1$0.40 / hr real-time; $0.24 / hr filesReal-time or files56, switches mid-sentenceSource
Soniox stt-rt-v5~$0.12 / hr real-time; ~$0.10 / hr asyncReal-time streaming60+ (translation included)Source
Gladia Solaria-3€0.16 / hr async (real-time €0.68 / hr)Async (files)100+Source

See all voice models · Latency and prices as each vendor publishes them.

Prices from vendors' official pricing pages, checked October 14, 2026. Not sponsored; order isn't a ranking.