ElevenLabs Agents bundles its own speech models and turn-taking into one per-minute price, $0.08 a minute past your plan's included minutes, with the language model on top. Vapi charges $0.05 a minute for hosting and passes each model provider through at cost. Dograh, our open-source AI voice agent platform, runs on your own servers or on Dograh Cloud.
Key Takeaways
- ElevenLabs Agents bills $0.08 a minute with its own speech models included.
- Vapi charges $0.05 a minute to host the call, with models at cost.
- An example 5-minute call is about $0.40 on ElevenLabs, $0.41 on Vapi, before the model.
Both products put a voice agent on the phone, and they split the bill in different ways. We compare setup, voices, telephony, pricing with one call worked line by line, and control over models and data, then cover the open-source option in the last section.
ElevenLabs Agents and Vapi at a glance
ElevenLabs sells a voice company's own stack as an agent product, while Vapi sells the plumbing that connects other companies' models. The table sums up each vendor's own pages.
| Row | ElevenLabs Agents | Vapi |
|---|---|---|
| What you buy | Agent minutes with ElevenLabs' speech-to-text, voices and turn-taking inside | Hosting for the call, with each model billed at the provider's cost |
| Price | $0.08 a minute past plan minutes, plans from $0 to $990 a month | $0.05 a minute hosting, plans from $0 to Premier |
| Language model | Supported LLMs or your own custom model, billed separately | Any provider, or your own custom model, at cost |
| Voices | ElevenLabs' 10,000+ voices | Any voice provider you connect, ElevenLabs included |
| Phone | Native Twilio, SIP trunk, batch calls. Carrier at cost | Vapi numbers and SIP at no Vapi charge, or Twilio, Vonage and Telnyx |
| Concurrent calls | 4 to 40 by plan, burst up to 3x | 4 to 30 by plan, $10 a month per extra line |
| HIPAA | BAA on Enterprise | HIPAA add-on at $2,000 a month |
| Hosting | ElevenLabs' cloud, or private deployment in your AWS or GCP account through sales | Vapi's cloud, no on-premise option |
Setup: a bundled stack or providers you pick
ElevenLabs Agents is the faster start if you want one vendor, and Vapi is the more open start if you want to choose each part. ElevenLabs' Agents overview describes an agent built from its own text to speech and speech recognition, your choice of a supported language model or a custom one, and a visual workflow builder for multi-step flows. Tests and evals sit in the same dashboard.
Vapi starts from the other end. You pick each provider, from speech-to-text to the voice, and Vapi runs the call between them. For multi-step calls it uses Squads, groups of assistants that hand calls to each other, since Vapi retired its older Workflows feature on 18 August 2026.
An ElevenLabs agent needs fewer decisions before it can talk, because the speech parts are already chosen. A Vapi agent takes a few more decisions, and in return you can swap any provider later without changing platforms. Developers who already have favourite providers tend to prefer that, and teams without a voice engineer tend to prefer the bundle.
Voices and speech-to-text
Voice quality is ElevenLabs' home ground, and Vapi can borrow it. ElevenLabs Agents speaks with ElevenLabs' own voices, more than 10,000 of them, plus voice clones you create on your plan. It also listens with ElevenLabs' own speech-to-text, Scribe v2 Realtime, so both ends of the audio come from one vendor.
On Vapi, ElevenLabs is one voice provider among several. You connect it with your own ElevenLabs account and pay ElevenLabs' API rate for the characters spoken, alongside the speech-to-text provider you prefer, such as Deepgram. The voice can sound the same on both platforms, but the bill and the account sit in different places.
That split matters when you tune for latency or accents. On ElevenLabs you tune inside one stack. On Vapi you can pair a voice with whichever speech-to-text model handles your callers' accents best, and test the pair on a real phone line before you commit.
Phone numbers, SIP and transfers
Both connect to the phone network, and both leave the carrier bill on your carrier's rates. ElevenLabs Agents offers a native Twilio integration for inbound calls on numbers you bring and outbound calls from verified caller IDs. A SIP trunk option connects existing phone systems, and batch calls place many outbound calls at once. ElevenLabs adds no telephony fee of its own. Its transfer tool supports conference and blind transfers, and SIP REFER on trunks.
Vapi sells its own numbers on every plan and lists Vapi Telephony and SIP at no charge. You can also connect Twilio, Vonage or Telnyx, and those carriers bill their own rates. The No Success Package includes 1 number, Core 5 and Pro 10.
If you already run a contact center or a PBX, both offer SIP. If you are starting from nothing, Vapi's built-in numbers save a carrier account, while ElevenLabs works through your own Twilio account or a SIP trunk.

Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
What each plan costs
ElevenLabs sells minutes in monthly plans, and Vapi sells a flat hosting fee with optional packages. ElevenLabs Agents comes in these plans.
| Plan | Price a month | Included call minutes | Extra minute | Concurrent calls |
|---|---|---|---|---|
| Free | $0 | 15 | $0.08 | 4 |
| Starter | $6 ($1 first month) | 75 | $0.08 | 6 |
| Creator | $22 ($11 first month) | 275 | $0.08 | 10 |
| Pro | $99 | 1,238 | $0.08 | 20 |
| Scale | $299 | 3,738 | $0.08 | 30 |
| Business | $990 | 12,375 | $0.08 | 40 |
| Enterprise | Custom | Custom | Custom | Higher limits |
Paid plans and extra minutes cost about the same per minute, $0.08, so the plan you choose mostly sets how many calls can run at once. Calls above your concurrency limit can burst to three times the limit at $0.16 a minute. The language model is billed separately from your ElevenLabs credits, at a rate that depends on the model.
Vapi charges $0.05 a minute for hosting, paid in advance, with $5 of free credits. The No Success Package costs nothing and allows 4 calls at once. Core is $29 a month for 10, and Pro is 10% of your hosting fees with a $999 monthly minimum, for 30. Extra lines cost $10 each a month, and the add-on for the Health Insurance Portability and Accountability Act (HIPAA) is $2,000 a month.
For plan-by-plan detail, our ElevenLabs pricing breakdown and Vapi pricing breakdown cover each vendor's plans and extras.
One 5-minute call, priced line by line
A worked example shows how the two models compare on a real call. Take one 5-minute US support call, past your ElevenLabs plan's included minutes. On Vapi, use Deepgram Nova-3 for speech-to-text at $0.0077 a minute, counted on both sides of the call as Vapi's own cost estimate does, and ElevenLabs Flash for the voice at $0.04 per 1,000 characters, about 400 characters a minute, so roughly $0.016 a minute. The Dograh Cloud column uses the same two providers with your own keys, with speech-to-text on the caller's audio, since the agent's own words are already text.
| Line | ElevenLabs Agents | Vapi | Dograh Cloud |
|---|---|---|---|
| Agent, hosting or platform fee | 5 x $0.08 = $0.40 | 5 x $0.05 = $0.25 | 5 x $0.01 = $0.05 |
| Speech-to-text | Included | 2 x $0.0385 = $0.077 | $0.0385 |
| Voice | Included | about $0.08 | about $0.08 |
| Total before model and phone | $0.40 | about $0.41 | about $0.17 |
| Language model | Billed from your credits | At the provider's cost | Your provider, your key |
| Phone line | Your carrier | Vapi number or your carrier | Your carrier |
So the two hosted options land close on this call, and the gap depends on your choices. A different speech-to-text model or a pre-recorded line moves the Vapi total, while ElevenLabs' price stays the same whatever you do inside its stack. The language model and the phone line are left out of the totals because both platforms treat them as pass-through costs.
Join the Dograh Community
Dograh is an OSS alternative to Vapi. Join our Slack community for queries, releases, best practices & community interactions.
Control over models, data and hosting
Vapi gives you more say over models, and ElevenLabs gives you more say over where the agent runs. Vapi's FAQ says you can bring your own custom models for any part of the pipeline, and also that Vapi does not support on-premise deployments. Your calls run in Vapi's cloud.
ElevenLabs keeps its own speech models in the agent but lets you choose the language model or plug in your own. For hosting, its private deployment runs the full ElevenAgents runtime inside your own AWS or GCP account, connected to your own model and telephony, and it is arranged through ElevenLabs' sales team.
Compliance follows the same pattern. ElevenLabs offers a HIPAA business associate agreement (BAA) on Enterprise. Vapi offers HIPAA as a $2,000-a-month add-on. Data residency outside the default region is an Enterprise item on ElevenLabs, so ask either vendor where your recordings and transcripts are stored before you sign.
Where an open-source platform fits
If you like ElevenLabs' voices but want to own the platform, Dograh is the open-source option. It runs the call flow, tools and telephony, and you choose the models. ElevenLabs voices and its Scribe v2 Realtime speech-to-text both run inside Dograh with your own ElevenLabs key, as does any other LLM, speech-to-text or voice provider, or open-weight models on your own servers.
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. That fee is the $0.05 line in the worked call above. You can self-host it with Docker or start on the managed cloud and move later.
The voice line can shrink further. Pre-recorded clips play a real human recording when one fits, with an ElevenLabs clone of the same voice as the fallback.
Every call also has a trace showing the speech-to-text, AI and voice time for each turn, and built-in QA scores each step of test and live calls.
For the wider field around each product, see our guides to Vapi alternatives and ElevenLabs alternatives.
Glossary
- Burst
- ElevenLabs' rate for calls above a plan's concurrency limit, $0.16 a minute, double the usual rate. Bursting allows up to three times the plan's limit, so steady overflow is better handled with a bigger plan.
- Hosting fee
- Vapi's charge for running the call itself, $0.05 a minute, separate from the speech-to-text, language model and voice providers that Vapi passes through at cost.
- Pass-through cost
- A platform billing a provider's service at the provider's own rate with no markup. Vapi passes models through at cost, and ElevenLabs passes the language model through on top of its agent minutes.
- Concurrent calls
- How many calls an agent can handle at the same moment. ElevenLabs Agents allows 4 to 40 by plan, and Vapi 4 to 30, with extra lines at $10 a month each.

