Cozzo
Find yours

Speech-to-Text & Text-to-Speech

Deepgram logoGladia logo

Deepgram vs Gladia: speech-to-text for voice agents and recorded calls

Pricing and features, side by side. Last verified October 27, 2026. Spot a wrong price? support@cozzo.io

Cozzo’s tip: start with the billing row. It’s where the surprises hide.

Cozzo’s verdict. Short answer: pick Deepgram for a live phone agent, because Flux English is $0.0065 a minute with turn detection built in and billing is per second. Pick Gladia for lots of recorded calls in English, French, German, Spanish or Italian (Solaria-3 files from €0.16 an hour), or when callers switch languages mid-call across Gladia’s 100. Cozzo’s pick for most English voice agents: Deepgram Flux.

5 of 8 rows differ

Deepgram and Gladia compared: type, billing style, starting price, billing, hipaa / baa, best for, key features, last verified
DetailDeepgram logoDeepgramGladia logoGladia
TypeSameBothSpeech-to-Text & Text-to-Speech
Billing styleSameBothPay as you go
Starting priceDiffersNova-3 streaming $0.0048/min ($0.29/hr); $200 free creditFiles from €0.16/hr (Solaria-3); real-time €0.68/hr
BillingDiffersPay as you go, billed per secondPay as you go, per hour (prepaid credits, billed in euros)
HIPAA / BAADiffersHIPAA: BAA for Enterprise customers; SOC 2 Type 2 (deepgram.com/pricing, checked Oct 22, 2026)GDPR, HIPAA, SOC 2 Type II
Best forDiffersStreaming speech-to-text for voice agentsReal-time multilingual transcription
Key featuresDiffers
  • Flux model with built-in turn detection
  • Per-second billing, no rounding
  • $200 free credit
  • 100 languages with automatic language detection and code switching (Solaria-1, live and files); Solaria-3 covers EN, FR, DE, ES, IT for files
  • Speaker diarization (files only) and custom vocabulary included
  • €50 free credits, valid for one year
Last verifiedSameBothOctober 27, 2026

Deepgram Pricing Explained →

Deepgram review: pricing, features and alternatives →Gladia review: pricing, features and alternatives →

Both sell speech-to-text APIs that voice agent builders plug into LiveKit, Pipecat or Vapi. Deepgram bills in dollars per second: Flux English $0.0065/min and Nova-3 $0.0048/min on live audio (both limited-time promotional rates; regular $0.0077), and Nova-3 recorded files at $0.0043/min ($0.26 an hour), with $200 in free credit. Gladia bills in euros per hour from prepaid credits: real-time on Solaria-1 is €0.68 an hour (€0.0113/min), files are €0.16 an hour on Solaria-3 (five European languages) or €0.27 on Solaria-1 (100 languages), falling with monthly volume, with €50 of free credit valid for a year. Deepgram also sells Aura-2 voices and an all-in Voice Agent API; Gladia is speech-to-text only.

Pick Deepgram if you’re building a live voice agent: lower live rates, turn detection inside Flux, per-second billing, more live streams at once (150 vs 30), plus voices and an all-in agent API from the same vendor.
Pick Gladia if most of your audio is recorded calls in its Solaria-3 languages, if callers mix languages in one call, or if you want HIPAA on self-serve without an Enterprise contract.

Sources: Deepgram · Gladia

Deepgram vs Gladia: API prices

Deepgram

Free to start
$200 credit, no card, no expiry
Live (real-time) transcription
Flux English $0.0065/min with turn detection; Nova-3 $0.0048/min (both promotional; regular $0.0077)
Live, multilingual
Flux Multilingual (10 languages) $0.0078/min; Nova-3 Multilingual $0.0058/min (regular $0.0092)
Recorded files
Nova-3 $0.0043/min ($0.26/hr); Multilingual $0.0052/min
Volume discounts
Growth: $4K+/year prepaid, up to ~20% off (Flux $0.0057, Nova-3 $0.0042)
Speaker labels
Live +$0.0020/min; included on files
Redaction
+$0.0020/min, live or files
Text-to-speech
Aura-2 $0.030 per 1K characters (7 languages)
All-in voice agent
Voice Agent API $0.075/min (STT, LLM and voice)
Concurrency (self-serve)
150 live streams, 50 REST requests
Billing
Per second, no rounding
HIPAA
BAA for Enterprise customers
Your audio
No training opt-in mentioned on the pricing page

Gladia

Free to start
€50 credit, valid one year (Gladia says about 73 hours of real-time)
Live (real-time) transcription
Solaria-1: €0.68/hr (€0.0113/min). No volume table for real-time
Live, multilingual
Same €0.68/hr, 100 languages with code switching
Recorded files
Solaria-3 from €0.16/hr (5 languages); Solaria-1 from €0.27/hr (100 languages)
Volume discounts
Automatic on files: down to €0.12/hr (Solaria-3) and €0.21 (Solaria-1) above 10,000 hrs a month. Top-up bonuses of 10% to 20% from €1,000
Speaker labels
Files only (not on live audio)
Redaction
PII redaction €0.07/hr, files only
Text-to-speech
None
All-in voice agent
None
Concurrency (self-serve)
30 live sessions, 25 file jobs
Billing
Per hour of audio from prepaid euro credits
HIPAA
GDPR, HIPAA, SOC 2 Type II listed on self-serve
Your audio
Self-serve is opted in to model training by default; free opt-out

Pay as you go list prices from deepgram.com/pricing and gladia.io/pricing, read Oct 27, 2026. Deepgram marks its Flux English and Nova-3 streaming rates as limited-time promotions. Euro prices are shown as published; Cozzo doesn’t convert them.

Worked example: one month of a busy voice agent

Deepgram

Assumptions
10,000 live call minutes a month (about 167 hours), English
Live transcription
Flux English: 10,000 × $0.0065 = $65 ($77 at the regular rate). Nova-3: $48
Plus 500 hours of recorded calls
Nova-3 files: 500 × $0.26 = $130
Total
$195 with Flux, $178 with Nova-3
Free credit covers
About 30,000 Flux minutes ($200 ÷ $0.0065)

Gladia

Assumptions
Same
Live transcription
Solaria-1 real-time: 166.7 hrs × €0.68 ≈ €113
Plus 500 hours of recorded calls
Solaria-3 files: 500 × €0.15 = €75 (Gladia’s own calculator figure)
Total
About €188
Free credit covers
About 73 hours of real-time, per Gladia

Arithmetic on pay as you go list prices only. Your LLM, voice and phone line are extra. Growth discounts, Gladia top-up bonuses and free credits would lower these totals.

Questions

Is Deepgram or Gladia better for voice agents?

For most English phone agents, Deepgram. Flux has turn detection built into the model, live audio costs $0.0065 a minute against Gladia’s €0.0113, and billing is per second. Gladia makes sense when callers speak many languages or switch mid-call, since Solaria-1 covers 100 languages with code switching.

Is Gladia cheaper than Deepgram?

For recorded files, yes on list prices: Gladia Solaria-3 starts at €0.16 an hour against $0.26 for Deepgram Nova-3, as long as the audio is English, French, German, Spanish or Italian. For live audio, no: Deepgram lists $0.0048 to $0.0065 a minute, Gladia €0.0113.

Which has more languages, Deepgram or Gladia?

Gladia’s Solaria-1 covers 100 languages, live and on files. Deepgram’s Nova-3 covers 45+, and Flux Multilingual, the model built for agents, covers 10 (English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian and Dutch).

Do Deepgram and Gladia have free tiers?

Deepgram gives $200 in free credit with no card and no expiry. Gladia gives €50, valid for one year, which it says is about 73 hours of real-time or 312 hours of Solaria-3 files.

Can I get a HIPAA BAA from Deepgram or Gladia?

Deepgram signs BAAs for Enterprise customers. Gladia lists GDPR, HIPAA and SOC 2 Type II on its self-serve plan. Either way, get the BAA signed before patient calls go through it.

Has Cozzo tested Deepgram or Gladia?

Not on their own. Cozzo’s phone tests measure whole voice agents. Deepgram Nova-3 was the listening model in Vapi’s default stack and in Cozzo’s own Pipecat build in published rounds; Gladia hasn’t been in a tested stack yet.