Retell suits phone teams. It prices each part of a call and includes 20 concurrent calls, and its business associate agreement is self-signed. ElevenLabs Agents suits teams wanting its voices and speech-to-text in one $0.08 minute, with that agreement on Enterprise. Dograh is the open-source AI voice agent platform we build, for your own servers or Dograh Cloud.
Key Takeaways
- Retell prices each part of the call, starting at $0.055 a minute.
- ElevenLabs Agents bundles its speech models into $0.08 a minute past plan minutes.
- Retell's business associate agreement is self-signed. ElevenLabs offers one on Enterprise.
Retell and ElevenLabs both sell hosted voice agents, and both let you use ElevenLabs voices. We compare them in one table, then on cost with one call worked line by line, on compliance and hosting, and on who each fits. The last section covers Dograh as the open-source option.
Retell AI and ElevenLabs Agents at a glance
Retell is a call platform that prices each part of the call, and ElevenLabs Agents is a voice company's own stack sold by the minute. The table uses each vendor's own pricing and compliance pages.
| Row | Retell AI | ElevenLabs Agents |
|---|---|---|
| How you pay | Per minute, by component: $0.055 voice infra, plus model, voice and phone | $0.08 a minute past plan minutes, language model on top |
| Plans | Pay as you go with $10 free credits, or Enterprise | Free to Business ($0 to $990 a month, 15 to 12,375 minutes), or Enterprise |
| Voices | Retell's own and other providers at $0.015 a minute, ElevenLabs from $0.04 | ElevenLabs' own voices, included in the minute |
| Language models | GPT, Claude and Gemini models priced per minute, plus speech-to-speech models | Supported LLMs or your own custom model |
| Phone | $0.015 a minute in the US via Twilio, $2 numbers, SIP trunking at no charge | Native Twilio or SIP trunk, carrier billed at cost |
| Concurrent calls | 20 included, $8 a month per extra line | 4 to 40 by plan, burst up to 3x at $0.16 a minute |
| HIPAA | BAA self-signed at no fee, custom BAA on Enterprise | BAA on Enterprise |
| Where it runs | Retell's cloud, with "Enterprise On-prem" on its homepage | ElevenLabs' cloud, or private deployment in your AWS or GCP account through sales |
Our ElevenLabs Agents vs Vapi comparison has the full ElevenLabs plan table, plan by plan.
How each one is built
Retell assembles a call from parts and shows you the price of each one, while ElevenLabs sells one stack it built itself. Retell starts every minute with a $0.055 voice infrastructure fee. On top of that you pick a language model, priced per minute, from GPT 5 mini at $0.008 to GPT 5.4 at $0.08. The voice and the phone line are priced per minute as well.
Retell also sells speech-to-speech models, where one model hears and speaks without separate speech-to-text and voice steps. GPT Realtime mini costs $0.07 a minute, and GPT Realtime $0.345.
ElevenLabs Agents keeps the listening and the speaking in-house. Scribe v2 Realtime handles the speech-to-text, ElevenLabs' own models do the voice, and the $0.08 minute covers both, along with turn-taking. You choose the language model from ElevenLabs' supported list or plug in your own, and it is billed separately from your credits.
So Retell gives you more ways to tune cost and model choice on one bill. ElevenLabs gives you fewer decisions and its own voice quality, since it builds the voices that Retell resells.
One 5-minute call on each
A worked example makes the two pricing models comparable. Take one 5-minute US call with the ElevenLabs Flash voice on both platforms and GPT 5 mini as the language model, past your ElevenLabs plan's included minutes. The Dograh Cloud column runs ElevenLabs Flash and Scribe v2 Realtime with your own ElevenLabs key, at ElevenLabs' API rates. Those are $0.04 per 1,000 characters for Flash, with about 400 characters to a minute of speech, and $0.39 an hour for Scribe v2 Realtime.
| Line | Retell AI | ElevenLabs Agents | Dograh Cloud |
|---|---|---|---|
| Platform or agent fee | 5 x $0.055 = $0.275 | 5 x $0.08 = $0.40 | 5 x $0.01 = $0.05 |
| Speech-to-text | No separate line | Included | about $0.0325 |
| Voice, ElevenLabs Flash | 5 x $0.04 = $0.20 | Included | about $0.08 |
| GPT 5 mini | 5 x $0.008 = $0.04 | Billed from your credits | Your OpenAI key |
| Phone line | 5 x $0.015 = $0.075 | Your carrier | Your carrier |
| Total shown | $0.59 | $0.40 plus model and carrier | about $0.16 plus model and carrier |
Two things move the Retell total. Switching to a Retell platform voice at $0.015 a minute brings the same call down to about $0.465. And that $0.59 already includes the model and the US phone line, which the other two columns leave to your own bills.
The ElevenLabs Flash voice costs $0.04 a minute inside Retell, about $0.016 a minute through the ElevenLabs API, and nothing extra inside ElevenLabs Agents, where it sits within the $0.08. Where you buy the same voice changes its price.
Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
Concurrency and add-ons
The per-minute rate is only part of the bill, because both vendors charge for capacity and extras. Retell includes 20 concurrent calls and charges $8 a month for each extra one, with Enterprise plans starting at 50 or more. Its add-ons are priced per minute or per call. A knowledge base or advanced denoising adds $0.005 a minute each, batch calls $0.005 per dial, and branded caller ID $0.10 per outbound call. AI Quality Assurance is free for the first 100 minutes, then $0.10 a minute.
ElevenLabs Agents ties concurrency to the plan, from 4 calls at once on Free to 40 on Business. Calls above the limit can burst to three times the limit at $0.16 a minute, double the usual rate.
For a busy outbound team, this row often matters more than the minute price. Twenty lines on pay as you go is generous for a pilot. On ElevenLabs, steady traffic above your plan's limit costs double through burst, so check whether a bigger plan beats bursting every day.
HIPAA, data and where each runs
Retell is the easier route to a business associate agreement (BAA) under the Health Insurance Portability and Accountability Act (HIPAA). Retell's compliance page lists SOC 2 Type 1 and Type 2, and lets you self-sign the BAA and the data processing agreement with no additional fee. A custom BAA sits on Enterprise. The same page says Retell does not currently operate services within the European Union.
ElevenLabs Agents offers the BAA on its Enterprise plan only. Data residency outside the default region and private deployment, which runs the full agent runtime in your own AWS or GCP account, are also arranged through ElevenLabs' sales team.
On hosting, Retell's homepage lists "Enterprise On-prem", which it describes as deploying Retell within your own infrastructure for strict security and data residency needs. That puts Retell's servers in your building, and the code stays closed. Our post on the open-source alternative to Retell AI explains why on-prem and open source are different things.
Join the Dograh Community
Dograh is an OSS alternative to Vapi. Join our Slack community for queries, releases, best practices & community interactions.
Who each one fits
Retell fits a phone operations team that wants to see and tune every part of the call. Per-component prices, 20 included lines, a self-signed BAA and outbound tools such as batch and branded calls suit a team running campaigns or a healthcare line on pay as you go. The trade is a bill with more lines, and add-ons such as QA that add to the minute.
ElevenLabs Agents fits a team that cares most about how the agent sounds and wants one vendor for both listening and speaking. The minute price is simple and stays the same whatever voice you pick. The trade is that the HIPAA agreement and private deployment both need an Enterprise contract.
Whichever you lean towards, test it on a real phone line before you decide. Most phone calls still reach the agent as 8 kHz audio, while a browser demo runs at 16 kHz, so an agent that sounds perfect in the dashboard can mishear callers on the phone.
A third option you can run yourself
If you want ElevenLabs' voices without either vendor's per-minute fee, or Retell's on-prem without closed code, Dograh is the open-source option. You pick the models, and the platform runs the call flow, tools and phone line around them. ElevenLabs voices and Scribe v2 Realtime speech-to-text both run inside Dograh with your own ElevenLabs key, and so does any other LLM, speech-to-text or voice provider. Speech-to-speech models from OpenAI and Grok run there too.
Self-hosted Dograh has no per-minute platform fee, but servers, model keys and phone minutes still cost money. Dograh Cloud charges a 1¢-a-minute platform fee. In the call above, that fee is the $0.05 line. You can self-host it with Docker, with the code under a BSD 2-Clause licence that you can read and change.
Quality checks are part of the platform. Dograh's QA scores each step of test and live calls, with tags, a 1 to 10 score, sentiment and a summary, and you choose what share of calls to score.
When a score looks off, the call trace shows the speech-to-text, AI and voice time for every turn.
For the wider field, see our guides to Retell AI alternatives and ElevenLabs alternatives.
Glossary
- Voice infrastructure fee
- Retell's base charge on every call minute, $0.055, before the language model, voice and phone line are added. It pays for running the call itself.
- Self-signed BAA
- A business associate agreement under HIPAA that a customer signs online without a sales contract. Retell offers one on pay as you go with no added fee. ElevenLabs offers its BAA on Enterprise.
- Speech-to-speech model
- One model that hears the caller and speaks back directly, instead of separate speech-to-text, language model and voice steps. Retell prices several per minute, and Dograh runs them from providers such as OpenAI and Grok.
- On-prem
- Running a vendor's software on servers inside your own infrastructure. It moves where the platform runs but not who owns the code, which is the difference from open-source software you can read and change.
