Cozzo
Find yours

Speechmatics Alternatives for Speech-to-Text Compared (2026)

Speechmatics lists Melia 1, its multilingual model (56 languages, switches language mid-sentence), at $0.40 an hour real-time and $0.24 an hour for files, Linden 1 for voice agents (built-in turn detection, real-time only) at $0.30 an hour, Oak 1 for medical audio at $0.30 an hour for files, and $100 in free credits with no card. Those are base rates: the page takes 33% off if you let Speechmatics train on your audio and up to 25% off with volume. The alternatives below: Deepgram Nova-3 streaming at $0.0048/min (about $0.29/hr, a promotional rate) or Flux, its voice-agent model, at $0.0065/min, with $200 in free credit; AssemblyAI Universal-Streaming at $0.15/hr and Universal-3.6 Pro Realtime at $0.45/hr; Soniox at about $0.12/hr real-time and $0.10/hr for files; and Gladia at €0.68/hr real-time and €0.16/hr for files, with €50 in free credits. Prices are from each vendor's own pricing page.

Speechmatics' pricing page switches its discounts on by default, so the first price you see is the lowest one. Turn both off to see the base rate before you compare.

Deepgram, AssemblyAI, Soniox and Gladia compared with Speechmatics on published speech-to-text pricing, free credits, languages and voice-agent models.

  • Deepgram logoDeepgram
    Starting price
    Nova-3 streaming $0.0048/min ($0.29/hr); $200 free credit
    Best for
    Streaming speech-to-text for voice agents
  • AssemblyAI logoAssemblyAI
    Starting price
    Streaming from $0.15/hr; $50 free credits
    Best for
    Low-cost streaming and batch transcription
  • Soniox logoSoniox
    Starting price
    ~$0.10/hr async; ~$0.12/hr real-time
    Best for
    Very low-cost speech-to-text
  • Gladia logoGladia
    Starting price
    Async from €0.16/hr; real-time €0.68/hr
    Best for
    Real-time multilingual transcription

STT models compared

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

What to check before you choose

  • Decide whether you'll trade your audio for a discount. Speechmatics' 33% model-training discount takes Melia 1 real-time from $0.40 to about $0.27 an hour. Gladia's self-serve plan trains on your audio by default, with an opt-out.
  • For voice agents, compare the turn-detection models: Speechmatics Linden 1 at $0.30/hr, Deepgram Flux at $0.0065/min (about $0.39/hr), and AssemblyAI Universal-3.6 Pro Realtime at $0.45/hr.
  • Check how streaming time is counted. AssemblyAI bills streaming on the whole time the connection is open, idle time included, so close sessions when calls end.
  • Count what's bundled. Soniox includes speaker diarization, language identification and formatting in its rate; AssemblyAI lists diarization at $0.12/hr extra; Gladia includes custom vocabulary, and diarization on files.
  • Medical audio? Speechmatics Oak 1 is $0.30/hr for files with real-time marked coming soon. ElevenLabs Scribe v2 Medical is $0.22/hr and HIPAA-eligible with BAAs. Get a BAA signed before any patient calls.
  • Mind currency and volume tiers. Gladia bills in euros from prepaid credits, and its Solaria-3 file rate falls from €0.16/hr (under 100 hours a month) to €0.12/hr above 10,000 hours.
  • Run your own call audio through the free credits: $100 at Speechmatics, $200 at Deepgram, $50 at AssemblyAI (per its billing docs) and €50 at Gladia. Published accuracy figures come from each vendor's own tests, not your callers.

Not sure which fits? Answer a few quick questions in the finder →

Deepgram vs AssemblyAI →Deepgram vs Speechmatics →Deepgram vs ElevenLabs →

Prices from vendors' official pricing pages, checked October 20, 2026. Not sponsored; order isn't a ranking.