Retell AI Pricing 2026: Per-Minute Costs and Add-Ons

In this article
As of October 2026, Retell AI charges $0.07 to $0.31 per minute for voice agents on its pay-as-you-go plan, with no platform fee and no minimum. Retell's own estimator defaults to $0.11 per minute before phone charges. Its Twilio-backed lines in the United States add $0.015 per minute, and each Retell number costs $2 a month. The first 20 simultaneous calls come with the account.
Everything else is metered on top. Retell bills extra capacity at $8 per call per month, add-ons by the minute and test calls at live rates (Retell pricing, Retell testing docs, as of October 2026). Its Enterprise plan is priced by quote, with custom capacity starting at 50+ calls.
Retell's rate card, line by line
Retell builds each minute from parts you choose. The table lists the parts a local business is most likely to use, from Retell's pricing page and phone number docs as of October 2026.
| Retell line item | Listed price | What it covers |
|---|---|---|
| Retell voice infrastructure | $0.055 per minute | The conversation engine, charged on every minute |
| Retell platform voices | $0.015 per minute | Text-to-speech; Cartesia, Fish, Minimax and OpenAI voices cost the same |
| Retell with ElevenLabs Flash | $0.040 per minute | A premium voice option |
| Retell with ElevenLabs Multilingual v2 | $0.06 per minute | A premium multilingual voice |
| Retell LLM: GPT 4.1 mini | $0.0128 per minute | A small, low-cost model |
| Retell LLM: GPT 5.1 | $0.04 per minute, or $0.08 on the fast tier | A mid-priced model |
| Retell LLM: Claude 4.5 Sonnet | $0.08 per minute | A larger model |
| Retell telephony on a Twilio line | $0.015 per minute | Calls carried on a Retell number |
| Retell telephony, your own SIP trunk | No Retell charge | Your carrier bills you instead |
| Retell local number, United States | Retell: $2.00 per month | Inbound and outbound calling |
| Retell toll-free number, United States | Retell: $5.00 per month | Toll-free inbound and outbound |
How one Retell minute adds up
Retell's estimator shows the default math. An LLM at $0.04, voice infrastructure at $0.055 and a platform voice at $0.015 make $0.11 a minute, with telephony at $0.00 because the default assumes your own carrier (Retell pricing, as of October 2026).
The published range follows from the same card. Retell's cheapest pieces, $0.055 infrastructure, a $0.015 voice and GPT 5 nano at $0.0016, come to about $0.07. Its priciest, an ElevenLabs v3 voice at $0.10 and GPT 5.5 on the fast tier at $0.32, push past $0.31 (Retell pricing, as of October 2026).
So the model and voice decide most of your rate. A clinic that wants a premium multilingual voice pays a different Retell price from a plumber happy with a standard one.
Retell's billing rules also shape the total (Retell pricing FAQ):
- Calls are tracked to the nearest second, with no rounding up per call.
- Silence and hold time are billed, because the speech engine keeps listening.
- After a transfer, the AI fee stops but the phone line keeps billing.
- Calls that fail to connect are not billed.
- On voicemail, billing runs only while the agent is active on the line.
The charges that sit outside the rate
These are the lines that make a Retell invoice larger than the per-minute math. The table is from Retell's pricing page and concurrency docs, as of October 2026.
| Retell charge | Listed price | When it applies |
|---|---|---|
| Retell extra concurrency | $8 per call line per month | Above the 20 lines included |
| Retell burst capacity | $0.10 per minute surcharge on the whole call | Calls started above your limit, if burst is on |
| Retell knowledge base | $0.005 per minute, and $8 per knowledge base per month after the first 10 | When an agent answers from your documents |
| Retell denoising or safety guardrails | $0.005 per minute each | When switched on |
| Retell PII removal | $0.01 per minute | When switched on |
| Retell AI quality assurance | $0.10 per minute after the first 100 minutes | Automated call review |
| Retell SMS | $0.01 per message, plus $20 per month | Texts sent from Retell numbers |
| Retell verified phone number | $10 per number per month | Outbound caller trust |
| Retell branded call | $0.10 per outbound call | Business name on caller ID |
Testing is the cost owners forget. Retell bills test calls at production rates and has no free sandbox. A batch of 20 test cases run three times bills 60 runs, and phone tests add the line charge (Retell testing docs).
Capacity has a cost of a different kind. When inbound calls hit your limit, Retell holds a new caller for about 40 seconds. If no line opens, the call goes to your fallback number, or ends if none is set (Retell concurrency docs).
Example: a month of Retell for an HVAC company
Example assumptions: a nine-technician HVAC company sends 600 inbound calls a month to a Retell agent. Each call lasts 2.5 minutes, for 1,500 minutes. The setup uses GPT 5.1, a Retell platform voice, a knowledge base, two Retell local numbers and 300 booking texts. Peak traffic stays under 20 calls at once. The team runs 60 minutes of phone tests. Prices are Retell's as of October 2026 (Retell pricing, Retell testing docs).
| Retell line | Math | Monthly cost |
|---|---|---|
| Retell minutes at $0.13 | $0.055 infrastructure + $0.04 GPT 5.1 + $0.015 voice + $0.015 phone line + $0.005 knowledge base, times 1,500 | $195.00 |
| Retell numbers | 2 at $2.00 | $4.00 |
| Retell SMS | 300 at $0.01, plus $20.00 | $23.00 |
| Retell test calls | 60 minutes at $0.13 | $7.80 |
| Retell total | $229.80 |
One change moves this a lot. On Retell, swapping the platform voice for ElevenLabs Multilingual v2 adds $0.045 a minute, or $67.50 for the month (Retell pricing, as of October 2026). A spike above 20 calls at once would, with burst switched on, add Retell's burst surcharge to every call started in it; with burst off, extra calls wait about 40 seconds and then go to your fallback number.
Outbound calls, do not call lists and STOP replies
Retell can also place outbound calls in batches and campaigns. Its Do-not-contact Handling is off by default. When on, the agent hears a stop request, confirms it once, ends the call and marks the number do not call (Retell DNC docs).
Two gaps matter for anyone running follow-up. Retell's batch calls and inbound calls do not check that flag. The call API checks it only when honor_internal_dnc is set to true. For texts, Retell treats a one-word reply such as STOP, UNSUBSCRIBE or QUIT as one of its opt-outs and marks the contact (Retell DNC docs).
The rules come from regulators, not from Retell. An FCC declaratory ruling released on February 8, 2024 treats AI-generated voices as artificial voices under the TCPA. So outbound AI calls need the called party's prior express consent unless an emergency purpose or exemption applies, and telemarketing calls must offer an opt-out (FCC Declaratory Ruling 24-17).
Telemarketers must also sync with the National Do Not Call Registry at least every 31 days. From October 1, 2026, the annual fee for registry data is $85 per area code after the first five (FTC Q&A on the Telemarketing Sales Rule). This is general information, not legal advice.
None of this compliance work appears on Retell's rate card. It belongs in your cost model all the same.
Why answer speed matters more than a cent a minute
A cent or two per minute is easy to compare. A lead that goes cold is not. In a Harvard Business Review (HBR) audit of 2,241 U.S. companies, only 37% answered a web lead within an hour, and 23% never answered at all (HBR, March 2011). Whatever platform you choose, price it against the calls it must catch after hours and on busy days.
Retell next to other pricing models
Retell is one of several ways to buy voice AI. The table uses each vendor's own page as of October 2026: Vapi, Bland AI, Synthflow and Goodcall.
| Vendor | Pricing model | Published price | Phone line |
|---|---|---|---|
| Vapi | Hosting fee plus provider pass-through | Vapi: $0.05 per minute hosting; its calculator shows $82–$129 for 1,000 minutes | Vapi's own telephony at no charge, or Twilio billed by Twilio |
| Bland AI | One bundled AI rate | Bland: $0.14 per minute with no platform fee, or $0.12 with a $299 monthly fee | Billed separately |
| Synthflow | Annual enterprise agreement | Synthflow: no public per-minute rate | Scoped in the contract |
| Goodcall | Per unique caller, no minute meter | Goodcall: $79 a month for 100 unique customers | A local number comes with every plan |
For the full market view, see the AI voice agent pricing comparison. For other platforms, see Synthflow alternatives and the Vapi voice agent explainer.
When a managed service is the better buy
Retell suits teams with a developer who will build and maintain the agent. A law firm or dental office without one pays in staff hours instead. Novacall AI is a managed option: it answers inbound calls and leads in under 60 seconds, 24/7/365, across voice, SMS, email and WhatsApp in 15+ languages. It qualifies callers with your questions, books appointments on your calendar and syncs to your CRM. Quotes are custom, based on call and lead volume, channels, integrations and workflow. See the side-by-side on the Retell alternative page.
Questions buyers ask about Retell pricing
Does Retell AI charge a monthly platform fee?
No. The pay-as-you-go plan has no platform fee and no minimum spend. You pay for the minutes your agents run, plus phone numbers, extra call lines and any add-ons you switch on. Enterprise customers negotiate their own pricing instead of using the public rate card.
Why is a Retell bill higher than the advertised per-minute rate?
The advertised range covers the AI part of a call. The invoice also carries the phone line, numbers, texts, knowledge base and other add-ons, burst capacity if it fires, and every test call. Silence and hold time are billed too, since the speech engine keeps listening.
Is Retell cheaper than Vapi or Bland AI?
It depends on the model and voice you pick. Retell and Vapi both pass model and voice costs through, so a small model with a basic voice keeps either one low. Bland folds the AI into one rate, which is simpler to budget but leaves less room to save through model choice.
What does Retell do with do not call requests and text opt-outs?
With Do-not-contact Handling switched on, the agent confirms the request, ends the call and flags the number. Single-word text replies such as STOP are logged as opt-outs. Batch calls and inbound calls ignore that flag, so outbound lists still need a check before each run.
What consent work does a Retell outbound campaign need?
Retell supplies tools such as do not call flags. Collecting consent, honoring each opt-out and checking the national registry are setup tasks for the business running the campaign, and a lawyer who works on telemarketing rules can confirm what your campaigns need.
Can a small business try Retell without spending money?
Only briefly. New accounts get a small sign-up credit, and after that every test message and test call bills at production rates. Run most checks as text simulations, which cost far less than voice minutes, and keep phone tests for final sign-off.
When does a managed service make more sense than Retell?
When nobody on staff will write call flows, connect the calendar and CRM, watch call quality and keep up with rule changes. Retell bills only for usage, so that labor falls to your team or a contractor. A managed provider folds the build and upkeep into one quoted fee.
Your next step
Start with your own numbers, not the rate card. Export last month's phone records into a spreadsheet. Add columns for minutes, the busiest hour of each weekday, weekend volume and anything that rang after closing. Note seasonal peaks too, such as the first heat wave or a holiday weekend, because those decide whether twenty lines are enough.
Price that workload on Retell with the model and voice you would really run, then add numbers, texts, tests and burst exposure. Set the finished figure beside a managed quote built from the same spreadsheet, and judge both against what a missed booking costs your shop.
