Comparison AI voice receptionists
Retell AI vs Vapi vs Synthflow (2026): What an AI Phone Agent Really Costs per Minute
Retell AI, Vapi, Synthflow and ElevenLabs Agents compared on published pricing: free credits, per-minute layers, a 1,000-minute worked example, and what Synthflow now sells. Checked 28 Oct 2026; not hands-on tested.

No one paid for this, and it has no affiliate links. How we review.
Research-based overview: we have not tested this hands-on yet, so there is no score. Facts come from the makers' own pages, linked below.
On this page
You can build an AI phone agent in an afternoon on some platforms. Whether it’s cheap to run is harder to see, because most of them price calls in layers. This page compares Retell AI, Vapi, Synthflow and ElevenLabs Agents on what they publish, then works through a 1,000-minute month. If you’d rather have someone else build it, jump to build or buy. If you only want a ready-made receptionist, our best AI receptionist roundup covers those.
A voice-agent platform is the plumbing that connects a phone line, speech-to-text (turning the caller’s voice into words), an AI language model that decides what to say, and a voice that says it.
Key takeaways
Verdict in one minute
- Retell AI and Vapi are the two you can sign up for and start testing today. Retell gives $10 in free credit and Vapi $5. Both bill per minute, in layers.
- Synthflow no longer has a self-serve plan for new customers. Its pricing page says “Enterprise contracts start at $30,000 annually”.
- ElevenLabs Agents is the self-serve third option, with monthly plans that include minutes and a free tier.
- On paper, Retell and Vapi land in a similar range for 1,000 minutes a month ($82 to $149 with the model and voice choices we name below). ElevenLabs Agents can’t be added to that number directly, because its plans bill the AI model and phone line separately. The differences are in the layers, concurrency (how many calls can run at once), compliance add-ons and phone line costs.
We haven’t placed test calls or used the free credits, so we can’t say which sounds better, answers faster or is easiest to set up. Those claims need hands-on tests.
Research-based comparison. We have not tested these products hands-on yet. Everything below comes from each company’s published pages, read on 28 October 2026 (one blog post on 26 October). Cost note: the per-minute totals are our own arithmetic from list prices and an AI model and voice we chose. Your real bill depends on your calls. Prices are US dollars before tax.
At a glance: AI phone agent platforms (USD, published prices)
| Tool | Free start | Platform price | Phone numbers / telephony | Self-serve signup |
|---|---|---|---|---|
| Retell AI | $10 free credit | Voice infrastructure $0.055 per minute, plus AI model, voice and phone line | Retell number $2 a month; your own telephony is free to connect; Twilio or Telnyx $0.015 per minute | Yes |
| Vapi | $5 free credit | Hosting $0.05 per minute, plus AI models at cost | SIP free; Twilio $0.008 in / $0.014 out per minute; Telnyx $0.0055 | Yes |
| Synthflow | None listed | Enterprise contracts start at $30,000 a year | Native telephony and SIP included in the contract | No, for new customers |
| ElevenLabs Agents | Free plan, 15 minutes | Plans from $6 a month; $0.08 per minute beyond included minutes | From your own provider | Yes |
What we checked
- Read the pricing pages of Retell AI, Vapi, Synthflow and ElevenLabs Agents on 28 Oct 2026
- Read Synthflow's billing and introduction docs on 28 Oct 2026
- Read ElevenLabs' help article and blog post on agent pricing (the blog on 26 Oct 2026)
- Worked out a 1,000-minute month for each provider from list prices (our arithmetic, naming the AI model and voice we assumed)
- Not done yet: placing test calls or trying any free credits
How per-minute pricing works: four layers
When a platform advertises “$0.05 a minute”, that’s usually only the first layer. A call’s cost has four parts:
- The platform fee. What the orchestrator charges: Retell’s voice infrastructure is $0.055 per minute, Vapi’s hosting is $0.05.
- The AI model. The language model that writes each reply. Retell lists its own per-minute prices for models (for example GPT 5.4 mini at $0.024 and Gemini 3.6 Flash at $0.024). Vapi passes model costs through at cost.
- The voice. The text-to-speech (TTS) engine. On Retell, the platform voice and several others cost $0.015 per minute; ElevenLabs voices cost $0.04 (Flash), $0.08 (v4 Turbo) or $0.10 (v4).
- The phone line. Twilio and Telnyx are two phone providers. Retell lists them at $0.015 per minute by default in the US. Vapi lists Twilio at $0.008 in and $0.014 out, and Telnyx at $0.0055.
Check the rules about what counts as a minute. Retell bills to the second, charges for silence and hold time, stops the AI fee once the call is transferred to a human, and doesn’t bill failed calls. For ElevenLabs Agents, billing is on connection time, with a 95 percent discount for silence longer than 10 seconds.
Retell AI
Retell’s page lists a $10 free credit and calls from $0.07 to $0.31 a minute, depending on the choices above. Chat agents start at $0.002 a message. Enterprise is custom.
- Concurrency: 20 concurrent calls are free; after that it’s $8 per extra concurrent line a month.
- Numbers: a Retell phone number is $2 a month, a verified number $10 a month; you can bring your own telephony at no charge.
- Add-ons: Knowledge Base is +$0.005 a minute (10 are free, then $8 each a month), PII removal +$0.01 a minute, a branded call +$0.10 per outbound call, AI quality assurance free for the first 100 minutes then $0.10 a minute, and SMS $0.01 a message.
- Compliance: the page lists SOC 2 and “HIPAA-ready”. It doesn’t say which plan includes a BAA, so that’s unconfirmed. For EU data protection terms, Retell mentions a custom data processing agreement on Enterprise.
Vapi
Vapi’s page gives $5 in free credit and a $0.05 per minute hosting fee, with the AI models, voice and speech-to-text passed through at cost.
- Concurrency: Vapi’s page lists 4 concurrent calls on pay-as-you-go, 10 on a $29 Core plan, and 30 on Pro, which has a $999 a month minimum (listed as 10 percent of hosting). Extra lines are $10 a line a month.
- Telephony: SIP is free; Twilio is $0.008 per minute inbound and $0.014 outbound; Telnyx $0.0055; Vonage $0.00814.
- Compliance: the HIPAA add-on is $2,000 a month and includes a DPA and BAA. Retention is 14, 30 or 180 days depending on plan.
- Its own estimate: Vapi’s cost calculator shows ranges for the layers: Deepgram speech-to-text $0.0095 to $0.0099 per minute, OpenAI $0.0077 to $0.0452, ElevenLabs voice $0.0146 to $0.0238. Vapi’s estimate for 1,000 minutes is $82 to $129 a month.
Synthflow: what changed
Synthflow’s pricing page now says “Enterprise contracts start at $30,000 annually”. Its docs describe “sales-led pricing”, say legacy workspaces may keep legacy billing, and say reselling was removed on 15 September 2026. We don’t use older third-party price lists, because they’re likely out of date.
What Synthflow says it sells is a platform built around what it calls the BELL framework (Build, Evaluate, Launch, Learn), plus a Test Center, WhatsApp, Salesforce and HubSpot integrations and its Aurora model. The contract covers native telephony and SIP, CRM integrations, an MSA and DPA (contract and data processing terms) and launch support. That’s a contracted enterprise product, not a self-serve tool, so we compare it only on what it publishes. For a small team, $30,000 a year is about $2,500 a month before any extras. We didn’t contact sales or book a demo.
ElevenLabs Agents: the self-serve option
ElevenLabs sells agents on monthly plans that include minutes, and the same plans cover its other products:
| Plan | Per month | Minutes included | Notes |
|---|---|---|---|
| Free | $0 | 15 | 4 concurrent; no commercial use |
| Starter | $6 | 75 | Commercial licence |
| Creator | $22 | 275 | |
| Pro | $99 | 1,238 | |
| Scale | $299 | 3,738 | 3 seats |
| Business | $990 | 12,375 | 10 seats |
Concurrency ranges from 4 to 40 calls across the plans. Beyond the included minutes it’s $0.08 per minute, and the AI model is billed separately. Bursting above your concurrency limit costs $0.16 per minute. Telephony comes from your own provider. ElevenLabs says prices were cut by up to 20 percent on 7 May 2026, and that BAAs are on Enterprise only. Prices exclude tax. For its newer voice agent features, see ElevenLabs’ realtime voice agents. ChatGPT’s own voice mode is a personal assistant rather than a phone agent; see ChatGPT Voice with GPT Live 1.1.
What we checked, and what we haven’t tested
We read all four pricing pages and the two Synthflow docs pages. We haven’t placed a test call, so we don’t compare how the voices sound, how fast they reply or how easy setup is. When we test, the plan is to use only the free credits, and call each agent with the same short script: a booking question, a transfer request and a silent pause. We’d record what appears on the usage screen, then compare it with our maths below.
Pros and cons
| Pros | Cons | |
|---|---|---|
| Retell AI | $10 free credit; clear published layers; 20 free concurrent calls | You pick the AI model and voice, which changes the price; per-country number prices unconfirmed |
| Vapi | $5 free credit; free SIP; pass-through model pricing | Totals depend on what you plug in; HIPAA add-on is $2,000 a month |
| Synthflow | Contract includes telephony, integrations and launch support | Enterprise contracts from $30,000 a year; no self-serve for new customers |
| ElevenLabs Agents | Fixed monthly plans with minutes included; free tier | Free has no commercial use; LLM and phone line billed separately |
1,000 minutes: our maths
Prices below are per month for 1,000 minutes of talk time, using list prices. We name the AI model and voice because the choice moves the total.
| Provider | Our maths | Monthly total |
|---|---|---|
| Retell AI | $0.055 platform + $0.015 voice + $0.024 AI model (GPT 5.4 mini) + $0.015 phone line = $0.109 a minute | About $109, plus $2 for a Retell number |
| Retell AI, pricier model | The same, with GPT 5.6 Terra at $0.064: $0.149 a minute | About $149 |
| Vapi | Vapi’s own estimate (Deepgram, OpenAI and ElevenLabs ranges above) | $82 to $129; add about $8 if you use your own Twilio inbound (1,000 × $0.008) |
| ElevenLabs Agents | Pro plan $99 with 1,238 minutes included, plus AI model and phone line | $99 plus AI model and telephony |
| ElevenLabs Agents, Creator | $22 + (1,000 − 275) × $0.08 = $22 + $58 | $80 plus AI model and telephony |
| Synthflow | $30,000 ÷ 12 | At least $2,500 a month |
Two notes. The Vapi and Retell figures include the model and voice; the ElevenLabs figures don’t, so don’t compare them directly. And all of this ignores tax, add-ons, failed-call retries, and the cost of building and maintaining the agent, which is mostly your time.
Outside the US
Retell lists phone numbers in many countries through Twilio, and in a smaller set through Telnyx. Check Retell’s phone-number page for your own country. Per-country rates aren’t shown on the page we read, so they’re unconfirmed. If your country isn’t listed, you can connect your own SIP phone line from a carrier you already use, and both Retell and Vapi can accept that. Vapi and ElevenLabs likewise rely on your own carrier or the providers named above. For GDPR (a data protection law), Retell mentions a custom DPA on Enterprise only. Check call recording and consent rules in your country before you take calls.
Build or buy?
Building on Retell or Vapi suits you if someone on the team is comfortable with APIs, webhooks (messages one system sends another when something happens) and call flows, and you want to control the prompts. Buying a finished receptionist suits you if you want it working next week. Our AI receptionist roundup compares those. A WhatsApp assistant may fit better if your customers message instead of call; see our WhatsApp AI chatbot guide. If you want to wire a voice agent into a calendar, CRM and follow-up emails without writing the glue yourself, you can get AIFix to build your AI phone agent. That’s our own service. Or do it yourself with a workflow tool; see n8n vs Make vs Zapier. Other platforms exist too, such as Bland AI, LiveKit and Voiceflow, but we haven’t checked them.
FAQ
Is Retell cheaper than Vapi?
Not clearly. Retell lists $0.055 for its platform and Vapi $0.05, but the model, voice and phone line change the total. Our 1,000-minute maths gives about $109 for Retell with the choices above, and Vapi estimates $82 to $129.
Does Synthflow have a free trial?
We found none on the pricing page. Synthflow lists enterprise contracts from $30,000 a year and no self-serve plan for new customers.
Do I pay during silence?
On Retell, yes: silence and hold time are billed. On ElevenLabs Agents, silence longer than 10 seconds gets a 95 percent discount. We haven’t confirmed Vapi’s rule, so check its live page.
Can I use my own Twilio number?
Retell says you can connect your own telephony for free; Vapi lists Twilio rates and free SIP. ElevenLabs Agents uses your own provider. Check each platform’s setup guide for how to connect it.
Is any of this HIPAA compliant?
Vapi sells a $2,000 a month HIPAA add-on with a DPA and BAA. Retell says “HIPAA-ready” without listing the plan with a BAA, and ElevenLabs offers BAAs on Enterprise. This isn’t legal advice, so ask your adviser before you take health information over the phone.
Which is best for people who don’t code?
We can’t say, because we haven’t tested setup. What the pages show: Retell and Vapi need you to choose models, voices and phone lines yourself, ElevenLabs Agents sells plans with minutes included, and Synthflow is a sales-led contract with launch support.
Sources
- Retell AI pricing, Retell AI (checked 28 Oct 2026)
- Vapi pricing, Vapi (checked 28 Oct 2026)
- Synthflow pricing, Synthflow (checked 28 Oct 2026)
- Billing, Synthflow docs (checked 28 Oct 2026)
- Synthflow documentation: introduction, Synthflow docs (checked 28 Oct 2026)
- ElevenAgents pricing, ElevenLabs (checked 28 Oct 2026)
- How much does ElevenAgents cost?, ElevenLabs Help Center (checked 28 Oct 2026)
- We've lowered API and agents pricing and introduced pay-as-you-go, ElevenLabs blog (checked 26 Oct 2026)
This post was drafted with help from an AI assistant, then checked, edited and approved by Numan Malik.



