Bottom line up front: Gladia is a speech-to-text API for developers who want transcription billed the simple way — per hour of audio, with no subscription and no seat fees. Its current self-serve rate starts at €0.16 per hour for batch transcription on its newest Solaria-3 model, new accounts get a €50 free credit, and diarization, language detection and custom vocabulary are included rather than sold as add-ons. If you are building a product that listens — a meeting assistant, a call analytics tool, a voice agent — it is one of the cheapest credible ways to start. If you need a finished app with a player and a team library, Gladia is not that product, and was never meant to be.
What is Gladia, and what does it do?
Gladia is a Paris-based speech-to-text and audio intelligence API, founded in 2022. It began as a production-hardened version of OpenAI’s Whisper and now runs its own Solaria model family. Solaria-1 is the broad-coverage model: 100+ languages, automatic language detection, and native code-switching, so a speaker who swaps languages mid-sentence stays in one continuous transcript. Solaria-3 is the newer model, tuned for noisy, fast, real-world business audio — call centers, sales calls and meetings — at a lower price.
You can send audio two ways. The async (batch) API transcribes recordings you upload or link, and the real-time API streams live audio and returns partial transcripts with sub-second latency — Gladia reports under 300ms for its streaming engine. Speaker diarization, word-level timestamps, enhanced punctuation and custom vocabulary are part of the core request, not paid extras. Optional audio-intelligence add-ons — summarization, sentiment analysis, translation and named entity recognition — are billed per hour on top when you enable them. We have reviewed the other developer-first options in this space, notably Deepgram and AssemblyAI; Gladia’s pitch against both is a lower headline rate with the intelligence features bundled.
How much does Gladia cost?
There are only two plans, and the self-serve one has no monthly fee at all. You fund a prepaid credit wallet, and hours of audio are drawn down against the balance.
| Plan | Async (batch) | Real-time | What you get |
|---|---|---|---|
| Self-serve | From €0.16/hr (Solaria-3) or €0.27/hr (Solaria-1) | €0.68/hr | €50 free credit for new accounts; 25 concurrent async / 30 real-time requests; 100+ languages, diarization and custom vocabulary included |
| Enterprise | Custom (annual) | Custom | Unlimited concurrency, custom models and fine-tuning, zero data retention, custom hosting, 99.9% uptime SLA, dedicated support |
The async rate falls automatically as your monthly volume grows — Solaria-3 steps down from €0.16 to €0.12 per hour at 10,000+ hours a month, with no commitment required to earn the discount. Large wallet top-ups earn bonus credits: +10% on €1,000, +15% on €2,500 and +20% on €5,000. Audio-intelligence add-ons cost €0.02–€0.07 per hour each when enabled.
What the pricing page doesn’t make obvious
Two things, both verified on Gladia’s current pricing page and its documentation of the July 2026 billing change. First, prices are now quoted in euros, and much of the older coverage quoting $0.61 an hour is out of date — that was the previous Starter plan’s dollar rate. Second, and more important if you are budgeting a small project: the old 10 free hours every month no longer exist. In July 2026 Gladia replaced the recurring free allowance with a one-time €50 signup credit, valid for one year — worth roughly 312 hours of async transcription on Solaria-3, 185 hours on Solaria-1, or about 73 hours of real-time. That is a generous runway, but once it is spent, a wallet top-up is the only way to keep transcribing. A side project that used to idle along free forever now has a meter running.
What are Gladia’s strengths?
- Genuinely cheap batch transcription. At €0.16 an hour on Solaria-3, an hour-long meeting costs about sixteen cents to transcribe — well under the per-hour rates we recorded for Deepgram and AssemblyAI in our own reviews.
- No subscription, no seats, no platform fee. The meter is the audio hour, which makes the bill trivial to forecast from one number.
- Multilingual audio is the home turf. 100+ languages, automatic detection and mid-sentence code-switching are core behaviour, not a premium tier.
- Features are bundled, not unbundled. Diarization, timestamps and custom vocabulary ride along in the base rate; rivals commonly charge per feature.
- Compliance basics are in place. GDPR, HIPAA and SOC 2 Type 2 are covered on self-serve, which matters for call and meeting audio.
What are Gladia’s weaknesses?
- API only — there is no product to log into. No player, no searchable team library, no no-code path. A non-developer who just wants meeting transcripts should use a finished app such as Sonix instead.
- The free allowance is one-time now. The €50 credit is large but finite and expires after a year; there is no recurring free tier to fall back on.
- Real-time costs over four times the batch rate (€0.68 vs €0.16 an hour), so a live-heavy workload changes the math completely.
- The strictest controls are Enterprise-only. Zero data retention, custom hosting and unlimited concurrency all sit behind a custom annual contract.
Who is Gladia for?
Developers and product teams adding transcription to something they are building: meeting assistants, call-center analytics, voice agents, media monitoring and captioning pipelines — especially teams whose audio is multilingual, noisy or code-switched, and European teams who want an EU-based, GDPR-covered provider. It is not for individuals who want an app, and it is not the cheapest possible option at extreme scale if you are willing to self-host an open model and run it yourself.
Frequently asked questions
Is Gladia free to try?
Yes. New self-serve accounts get a €50 credit, valid for one year — roughly 312 hours of batch transcription on Solaria-3. There is no longer a recurring monthly free allowance; the credit replaced it in July 2026.
How much does Gladia cost per hour?
Async transcription starts at €0.16 per hour on Solaria-3 and €0.27 on Solaria-1, falling to €0.12 and €0.21 at 10,000+ hours a month. Real-time streaming is €0.68 per hour. Audio-intelligence add-ons cost €0.02–€0.07 per hour each.
How is Gladia different from Deepgram or AssemblyAI?
All three are developer speech-to-text APIs. Gladia’s headline batch rate is lower than the list rates in our Deepgram and AssemblyAI reviews, diarization and vocabulary are included in the base rate rather than billed per feature, and its strongest suit is multilingual and code-switching audio.
Does Gladia have an app for transcribing my meetings?
No. Gladia is an API for builders. If you want a finished transcription product with a player and library, look at a tool like Sonix or one of the meeting assistants we have reviewed.
Our verdict
Gladia makes the simplest promise in speech-to-text: send audio, pay by the hour, get a good transcript — and at €0.16 an hour for batch work it keeps that promise cheaply, with the features rivals charge extra for already included. Just go in with the current facts, not last year’s: prices are in euros now, and the free tier is a one-time €50 credit, not free hours forever. For a developer building anything that listens, that is still an easy starting point to recommend.