ChatGPTTL;DR
- Retell's own headline is $0.07 to $0.31 a minute, pay as you go, with $10 in free credits and no monthly minimum.
- The platform fee is $0.055 a minute. A recommended model adds $0.064, a platform voice $0.015 and Retell telephony $0.015, so a sensible build lands near $0.15 a minute.
- A plumber taking 300 inbound minutes a month pays about $46 on Retell, $25 to $41 on Vapi and $190 on our flat-fee agent.
- Vapi's hosting fee is $0.05 a minute with the model, voice and transcriber billed at provider cost, which is why its bill is a range.
- Skip the speech-to-speech models at $0.345 to $0.38 a minute. They triple the bill for a receptionist job.
- The platform bill is the small number. The person who builds and maintains the agent is the real cost, and it is the one a flat fee covers.
Retell AI pricing looks like one number until you add the model, the voice and the phone line. Then it looks like a receipt with nine lines, and the receipt is what a home service business actually pays.
Disclosure: Skill Mammoth sells a done-for-you AI phone agent at $150 a month, and it competes with building your own on Retell or Vapi. We have no arrangement with either platform. Every price below was read from the vendor pricing pages on November 4, 2026.
This guide prices a Retell agent line by line, builds a 300-minute month for a plumbing company, does the same on Vapi and says where a flat fee wins. If you want the product comparison instead, the AI receptionist guide prices nine finished products.
Retell's $0.055 is the smallest line on the bill
Retell bills pay as you go, meaning you load credit and the balance drops per second of talk time, with no plan and no minimum. The pricing page says calls are tracked to the nearest second and summed at the end of the cycle, with no rounding up per call.
The platform fee is called Voice Infra and costs $0.055 a minute. It covers the real-time call engine and speech to text (STT, the part that turns the caller's voice into words). Everything else is an add-on you pick when you build the agent.
| Line | Per minute | What it is |
|---|---|---|
| Voice Infra | $0.055 | The platform fee. Includes STT and the call engine. Every call pays it. |
| LLM, recommended tier | $0.064 to $0.16 | The language model that decides what to say. GPT 5.6 Terra and Claude 5 Sonnet are $0.064, GPT 5.4 is $0.080, GPT 5.5 is $0.16. The cheapest listed model is GPT 5 nano at $0.0016, the dearest is GPT 6.1 Sol Ultrafast at $0.384. |
| Fast tier LLM | Double the standard rate | Lower latency on the same model. GPT 5.6 Terra goes from $0.064 to $0.128. |
| Text to speech (TTS) | $0.015 to $0.10 | The synthetic voice. Retell Platform Voices, Cartesia, OpenAI and Minimax are $0.015. ElevenLabs Flash is $0.04 and the ElevenLabs v3 and v4 voices are $0.08 to $0.10. |
| Telephony | $0.015 or $0 | Retell's Twilio and Telnyx lines are $0.015. Bring your own carrier through SIP trunking (plugging your existing phone provider into Retell) and the line is free. |
| Knowledge base | $0.005 | Charged on minutes where the agent reads your uploaded documents. Ten knowledge bases are free, then $8 a month each. |
| Other add-ons | $0.005 to $0.10 | Advanced denoising $0.005, safety guardrails $0.005, PII removal $0.01, AI quality assurance $0.10 after the first 100 minutes. |
| Verdict | Voice Infra plus a recommended model, a platform voice and Retell telephony is $0.149 a minute before add-ons. The model is the line that moves the bill, and it moves it by a factor of 240 between the cheapest and dearest option. | |
Retell's own calculator defaults to a cheaper build. It shows $0.11 a minute from a $0.04 model, the $0.055 platform fee and a $0.015 voice, with telephony at zero because it assumes you bring your own carrier.
The speech-to-speech models are the trap. GPT Realtime is $0.345 a minute and GPT Realtime 2.1 is $0.38, before the platform fee. They sound more natural on a demo. For a receptionist that books a drain cleaning, they triple the bill for a difference most callers will not notice.
The monthly lines that calculators leave out
Four items on Retell's page bill by the month, not the minute. They are small for a single-location shop, and one of them is printed two different ways.
- A Retell phone number is $2 a month. Porting your own number is free.
- Concurrency, meaning how many calls the agent can take at once, is free for the first 20 and $8 a month for each one after that.
- Knowledge bases are free for the first ten, then $8 a month each. One block on the page phrases this as "$80/month free value (10 KBs)."
- A verified phone number, which shows your business name on outbound caller ID, is printed as "$10.00/Phone number/month" in one block and "One-time $10 fee" in another. Ask Retell which one applies before relying on it.
- SMS is $20 a month plus $0.01 a message.
Enterprise is custom pricing with no public minute commitment. The FAQ says enterprise concurrency starts at 50 calls or more. A three-truck plumbing company does not need it.
What a plumber's 300-minute month costs on Retell, Vapi and a flat fee
Here is the arithmetic for a plumbing company taking 300 inbound minutes a month, which is roughly 100 calls at three minutes each. The Retell build uses a recommended model, a platform voice, Retell telephony and one knowledge base. The Vapi build uses the three provider examples printed on its page.
| Line | Retell, pay as you go | Vapi, No Success Package | Skill Mammoth |
|---|---|---|---|
| Platform fee | $0.055 Voice Infra | $0.05 hosting | Flat $150 a month |
| Speech to text | Included | $0.0095 to $0.0099 (Deepgram) | Included |
| Language model | $0.064 (GPT 5.6 Terra, standard) | $0.0077 to $0.0452 (OpenAI) | Included |
| Voice | $0.015 (Retell Platform Voices) | $0.0146 to $0.0238 (ElevenLabs) | Included |
| Telephony | $0.015 (Retell Twilio or Telnyx) | $0 (Vapi SIP) to $0.008 (Twilio inbound) | Included |
| Knowledge base | $0.005 | Not printed | Included |
| Per minute | $0.154 | $0.0818 to $0.1369 | 200 minutes included, then $0.40 |
| 300 minutes | 300 x $0.154 = $46.20 | 300 x $0.0818 to $0.1369 = $24.54 to $41.07 | $150 + (100 x $0.40) = $190 |
| Verdict | On the invoice, the platforms win by $144 to $165 a month. The Retell and Vapi totals exclude the phone number, the setup time and the person who builds and maintains the agent. The flat fee includes all three plus a $199 setup, and the gap is what you are paying that person. | ||
Three things the platform totals leave out:
- The phone number. Retell's is $2 a month. Vapi includes one number on the No Success Package and prints no price for more.
- The setup. Writing the prompt, connecting the calendar, testing the transfer rules and recording a greeting takes a technical person days, not hours, the first time.
- The upkeep. Models get deprecated, a voice gets renamed, the calendar integration breaks. Somebody has to notice at 7am on a Monday.
Our $150 agent is live on your number in about a week, month to month, with a $199 setup fee. Details are on the AI phone agent page. If you have a developer on payroll or you are one, the platforms are cheaper and the rest of this guide is for you.
Want the agent without the build?
Tell us your trade and your call volume. We will say whether a Retell build, a Vapi build or a flat fee is the cheaper path for your shop.
Vapi pricing, line by line
Vapi prices the same job a different way. Its hosting fee is $0.05 a minute, pre-paid, and every other line is a pass-through from the provider at cost with no markup. That is why Vapi's page prints ranges where Retell prints a number.
| Line | Price | Note |
|---|---|---|
| Vapi hosting | $0.05 a minute | Pre-paid. The only line Vapi keeps. |
| Transcriber (STT) | $0.0095 to $0.0099 | Deepgram, the printed example |
| Intelligence model (LLM) | $0.0077 to $0.0452 | OpenAI, the printed example |
| Voice (TTS) | $0.0146 to $0.0238 | ElevenLabs, the printed example |
| Transport | $0 to $0.014 | Vapi telephony, SIP and WebRTC are free. Twilio inbound $0.008, Twilio outbound $0.014, Telnyx $0.0055, Vonage $0.00814. |
| No Success Package | $0 to start | Pay as you go, $5 free credits, 1 phone number, 4 concurrent calls, 14 days of data retention |
| Core | $29 a month | Pre-paid and auto-renewing. 5 numbers, 10 concurrent calls, 30 days of retention, email support |
| Pro | 10 percent of hosting, $999 a month minimum | 10 numbers, 30 concurrent calls, 180 days of retention, a 99 percent uptime SLA |
| Add-ons | $10 to $2,000 a month | Extra concurrency $10 a line, extra organization $20, HIPAA compliance $2,000 |
| Verdict | Vapi's calculator at 1,000 minutes estimates $82 to $129 a month, which works out to about $0.08 to $0.13 a minute. Cheaper than Retell on paper, with four concurrent calls instead of twenty on the free tier. | |
Vapi's "No Success Package" is a real plan name. It means no onboarding help, no support SLA and a four-call concurrency cap. For a single phone line taking one call at a time that cap never matters. For a storm-season roofing office it does.
Line by line against Retell, the differences are small and they mostly favor Vapi on price and Retell on included capacity.
| Item | Retell | Vapi |
|---|---|---|
| Platform fee per minute | $0.055, STT included | $0.05, STT extra at about $0.01 |
| Model billing | Retell's own per-minute rate card | Provider cost, no markup |
| Telephony per minute | $0.015, or free with your own SIP trunk | Free on Vapi SIP, $0.008 Twilio inbound |
| Free credits | $10 | $5 |
| Free concurrent calls | 20, then $8 a month each | 4, then $10 a line a month |
| Phone number | $2 a month | 1 included, no price printed for more |
| Knowledge base | $0.005 a minute, 10 free then $8 a month | Not printed |
| Minimum | None | None on No Success Package, $999 on Pro |
| Verdict | Pick Vapi if you want to shop your own model and voice providers and pay their rates. Pick Retell if you want one rate card, twenty free concurrent calls and a published knowledge base price. Neither choice moves a small shop's bill by more than $25 a month. | |
The AI agents guide covers where a voice agent sits next to the email and quoting agents that most shops should build first.
What the third-party price guides get wrong
Search "retell ai pricing" and the top results are written by Retell's competitors. Two of them deserve a note, because readers quote them.
| Guide | Dated | What it prints | Against Retell's page today |
|---|---|---|---|
| Dialora (competitor) | Updated October 14, 2025 | GPT-4o and Claude 3.5 as the models, telephony at $0.01, a headline of $0.13 to $0.31 a minute, a 1,000-minute example at $105 | None of the model names or the telephony rate appear on the current page. The $105 example is close to today's math by accident. |
| Eesel (competitor) | Last edited October 5, 2026 | $0.055 platform fee, $0.015 voices, $2 and $10 phone numbers, $8 per extra concurrent call | Matches the vendor page on every line we checked. No worked monthly example, which is the one thing a buyer needs. |
| Verdict | Check the date before you quote a number. One of the two top guides is a year stale and still ranks. | ||
The lesson is simple. Retell changes its model list every few weeks, so a price guide without a read date is a guess. The date on ours is at the top.
Who should build on Retell
Retell is the right platform for three kinds of buyer.
- An agency building agents for several clients. Twenty free concurrent calls, a published knowledge base price and a per-second meter make the margin easy to compute. The AI-first contractor playbook shows the layer a phone agent sits in.
- A shop with a developer on staff. If someone already maintains your Jobber or ServiceTitan integrations, a Retell agent is a few days of work and $46 a month at 300 minutes.
- A high-volume operator. Past 2,000 minutes a month the gap between $0.15 and a flat fee with a $0.40 overage becomes real money, and at that volume you can afford the person to run it.
Everyone else should read the two monthly numbers again. At 300 minutes the platforms save $144 to $165 a month against a done-for-you agent. That is less than one missed emergency call, and the missed-call guide puts a number on what those cost.
Three things to skip
- Speech-to-speech models for a receptionist. GPT Realtime at $0.345 a minute is more than six times the platform fee. A $0.064 model with a $0.015 voice books the same job.
- The fast tier by default. It doubles the model line. Turn it on only if a test caller complains about the pause, and test with the standard tier first.
- ElevenLabs v3 or v4 voices on a budget build. At $0.08 to $0.10 a minute the voice costs more than the model. The $0.015 platform voices are what most branded AI receptionists ship with anyway.
Not sure the phone is your bottleneck?
Send us your website. The 10-minute audit shows where calls and form leads are leaking before you spend a dollar on an agent, within 48 hours.
Frequently asked questions
How much does Retell AI cost per minute?
Retell's pricing page quotes $0.07 to $0.31 a minute for voice agents, read November 4, 2026. The platform fee is $0.055 a minute, and you add a model (recommended models are $0.064 to $0.16), a voice ($0.015 to $0.10) and telephony ($0.015, or free with your own carrier). A typical build with a recommended model and a platform voice costs about $0.15 a minute.
Does Retell AI have a monthly fee or minimum?
No. Retell is pay as you go with $10 in free credits and no minimum. The monthly items are optional: a Retell phone number at $2, SMS at $20, concurrency beyond 20 calls at $8 each and knowledge bases beyond ten at $8 each. Enterprise plans are custom priced and start at 50 or more concurrent calls.
How does Vapi pricing compare to Retell?
Vapi charges $0.05 a minute for hosting and passes the model, voice and transcriber through at provider cost, so its printed example works out to about $0.08 to $0.13 a minute. Retell is $0.055 plus its own rate card, about $0.15 for a recommended build. Vapi gives 4 free concurrent calls and $5 in credits, against 20 and $10 on Retell.
What does a Retell AI agent cost a small business per month?
For a plumbing company taking 300 inbound minutes a month, a Retell agent with a recommended model, a platform voice, Retell telephony and one knowledge base costs about $46.20, plus $2 for a phone number. The same minutes on Vapi run $24.54 to $41.07. Neither figure includes the time to build and maintain the agent.
Is a done-for-you AI phone agent cheaper than building on Retell?
Not on the invoice. Our agent is $150 a month with 200 minutes included, then $0.40 a minute, so 300 minutes is $190 against about $46 on Retell. The flat fee includes the build, the calendar integration, the number and the upkeep. It is cheaper only when you count the hours a technical person would spend doing those things.
Does Retell charge for failed or unanswered calls?
Retell's pricing page says each call is tracked to the nearest second and billed on total accumulated minutes with no rounding up per call, so a call that never connects accrues no talk time. Its billing docs point to a separate page of exceptions to per-minute pricing that was not reachable on November 4, 2026. Ask support for that list before launching an outbound campaign.