How Much Does Retell AI Cost in 2026? Per-Minute Pricing and Real Monthly Bills

Retell AI charges $0.07 to $0.31 a minute for voice agents with no platform fee, and a realistic support call (GPT 5.4 mini, Retell's own voice, US telephony and a knowledge base) comes to $0.114 a minute, or $1,144 a month at 10,000 minutes. Chat agents bill per AI message from $0.0005, new accounts get $10 of free credit, and Enterprise is custom-priced.

Prices checked September 19, 2026 on retellai.com/pricing; the headline range, the $0.055 voice infrastructure rate, the $10 credit and the 20 free concurrent calls were rechecked unchanged on September 24, 2026.

For the product overview, read our Retell AI complete guide. This page stays on what Retell will cost you, and why.

What goes into a Retell minute?

Retell sells a pipeline, and each stage has its own meter. Every voice minute carries four charges, and the published range comes from adding them up.

The bottom of the headline range, $0.07, is exactly the voice infrastructure plus the cheapest voice. It leaves out the model and the phone line. The cheapest configuration that actually answers a phone call on a Retell number is $0.055 + $0.015 + $0.0016 + $0.015, or $0.0866 a minute. Bring your own SIP trunk and Retell's part drops to $0.0716, though your carrier then bills you directly.

The model is where the range lives. Retell lists more than 25 models, and the per-minute price spans GPT 5 nano at $0.0016 up to GPT 6 Astra at $0.32, a 200-fold difference before fast tier. Retell's four "Recommended" models sit between $0.064 (GPT 5.6 Terra, Claude 5 Sonnet) and $0.16 (GPT 5.5). Fast tier routes calls through priority capacity and, in Retell's own words, is priced at "1.5x the standard rate". The price table matches that for the GPT 4.1 models but shows exactly double for the GPT 5 family (GPT 5.4 is $0.080 standard and $0.16 fast). Budget on the table's figure.

There is also a speech-to-speech mode, priced on its own tab of the component table: OpenAI's realtime models at $0.07 a minute for the mini versions and $0.345 to $0.38 for the full ones. The tab doesn't say whether the $0.055 voice infrastructure is added on top, so ask before you model it.

What does Retell's calculator say a support line costs?

Retell's pricing page has an estimator. We set it to 10,000 minutes a month with GPT 5.4 mini (a mid-priced model that handles tool calls), Retell's own platform voices and Retell's Twilio/Telnyx telephony. It returned $0.109 a minute and $1,090 a month. The calculator prices its "Retell Twilio/Telnyx" option at $0.015 a minute, while Retell's international-calling docs list Telnyx US calls at $0.03, so a Telnyx number may cost $0.015 a minute more than the calculator shows. Ask which carrier your numbers are on.

The calculator leaves out the add-ons and the monthly items. All five call add-ons, Knowledge Base included, show as selected chips in the run above, and the add-ons line still reads $0.000 a minute, so a support agent that answers from your docs needs another $0.005 a minute on top. Phone numbers, concurrency and verified numbers are monthly subscription items, not per-minute charges, so the calculator never shows them. Our "support stack" below is therefore the calculator's $0.109 plus the knowledge base: $0.114 a minute.

Which Retell add-ons and monthly items cost extra?

Retell prices optional features per minute or per call, and a further layer as monthly subscriptions.

Per-minute and per-call add-ons:

• Knowledge base: +$0.005 a minute.
• Advanced denoising: +$0.005 a minute.
• Safety guardrails: +$0.005 a minute.
• PII removal: +$0.01 a minute.
• AI quality assurance: $0.10 a minute after the first 100 free minutes. Retell's AI QA docs let you set a sampling rate, which is how you keep this from doubling a bill.
• Batch call: +$0.005 per dial. Branded call: +$0.10 per outbound call. SMS: $0.01 a message.

Monthly items, billed to your card at the end of each cycle rather than drawn from credits:

• Retell phone number: $2 a month (no charge for numbers you import). US toll-free numbers are $5 a month, per Retell's number docs, and inbound toll-free minutes bill at $0.06.
• Retell SMS: $20 a month.
• Concurrency: the first 20 simultaneous calls are free, then $8 per extra call a month.
• Knowledge bases: the first 10 are free, then $8 each a month.
• Verified phone number: $10 a month.

What does Retell cost at 1,000, 10,000 and 50,000 minutes?

Here is the arithmetic at three sizes, using list prices and two stacks. The support stack is our calculator configuration plus a knowledge base ($0.114 a minute). The premium stack swaps in GPT 5.4 and an ElevenLabs voice and adds denoising and PII removal: $0.055 + $0.040 + $0.080 + $0.015 + $0.005 + $0.005 + $0.010 = $0.210 a minute. We assume a 4-minute average call.

The signup credit takes $10 off the first month, the only discount on the published price. The large row is where you'd expect to talk to Retell's sales team, since the Enterprise plan advertises "volume pricing that scales as you grow". The pricing page gives no number, but Retell's own blog names a floor of "as low as $0.05 per minute", and competitor guides put $8,000 on Enterprise or on its managed setup (details in the Enterprise section below). Treat the large row as the list price you negotiate down from.

Going from the support stack to the premium one nearly doubles the bill, and $0.056 of the $0.096 difference is the model. The voice upgrade costs $0.025 and the two add-ons, denoising and PII removal, $0.015.

What do Retell chat agents cost for 3,000 conversations?

Retell's chat agents are priced per AI message, and the rate is set by the model. There's no separate platform charge per chat.

A support team handling 3,000 web-chat conversations a month, at 6 AI replies each, sends 18,000 AI messages:

The pricing card headlines chat at "$0.002+", but the table under it lists two models below that, so the cheapest month is $9. Both are small models: run your own tool calls and policy questions through them in the playground before you budget on $9. Chat over SMS adds $0.01 per message and the $20 monthly Retell SMS subscription. Text testing in the playground and in simulations bills at these same chat rates (see the next section).

What pushes a Retell bill up?

Retell documents its billing surprises: three in the pricing FAQ, the rest in its docs.

Long prompts are billed as longer calls. Retell's billing exceptions page says that when an agent's prompt passes 4,000 LLM tokens, billing scales by tokens ÷ 4,000. The count includes the global prompt, tool descriptions, node prompts, the transcript so far, tool results and retrieved knowledge-base content, so it grows as a call goes on. The worked example: a 60-second call at 4,800 tokens bills as 72 seconds.

The screenshot on the same docs page tells a slightly different story: it shows a 3,500-token threshold and applies the surcharge to the LLM line only ($0.050 a minute becomes $0.074 at 5,209 tokens). The two readings produce very different bills. Take our support stack at 10,000 minutes with a 6,000-token context. On the screenshot's reading, the factor is 6,000 ÷ 3,500 = 1.71 on the LLM line only, which adds 10,000 × $0.024 × 0.71 = about $171 a month. On the text's reading, the factor is 6,000 ÷ 4,000 = 1.5 on the whole billed duration, which adds 10,000 × $0.114 × 0.5 = $570. Ask Retell which one your account gets, and keep prompts short either way: the docs name "Flex mode" as a common trigger because it compiles every node into one context.

Silence and hold time are billed. Retell's pricing FAQ says billing covers the entire call because the speech-to-text engine keeps listening. An agent that puts callers on hold while a slow API responds pays for every second of it.

Transfers keep the phone line open. Once a call transfers to a human, the AI fee stops but telephony continues for the rest of the transferred call. That's small on a US number and large on some others.

International telephony. Retell's international calling table lists $0.015 a minute for the US but $0.06 for US toll-free, $0.10 for the UK, Germany, Spain and Australia, $0.28 for Japan and $0.80 for the Philippines. On the support stack, a UK call is $0.199 a minute, not $0.114.

Concurrency and bursts. Above 20 simultaneous calls you buy lines at $8 each a month. If you switch on concurrency burst instead, calls above your limit go through with a $0.10-a-minute surcharge on the entire call, not just the minutes spent over the limit. A call that would cost $0.114 a minute costs $0.214 if it started during a burst.

Testing is billed at production rates. Retell's testing pricing page says it directly: "There's no separate test tier and no free testing sandbox." Simulations bill two models (your agent and the simulated caller), batch tests bill every case on every run plus a grading unit, and web or phone test calls bill per minute. Conductor, the AI copilot that helps you build agents, is free for 30 messages per user and 200 per workspace a day, then costs $0.20 a message if an admin turns on pay-as-you-go; otherwise it pauses until midnight Pacific. One G2 reviewer's only complaint was that "the per-minute pricing structure can feel a bit steep when you are just starting out and doing a lot of testing."

Small charges that add up on outbound. Calls shorter than 10 seconds that use a dynamic opening message bill as 10 seconds. Branded caller ID is $0.10 per outbound call, which on a 1-minute outbound call nearly doubles the price.

What doesn't cost you: calls that fail to connect aren't billed, voicemail is billed only while the AI is active on the line, and minutes are tracked to the second with no per-call rounding.

How does Retell bill: prepaid credits or invoices?

New accounts run on prepaid credits. Per Retell's billing docs, usage (voice, LLM, telephony) is deducted from your balance in real time, and subscription items are charged to your card at the end of the month, prorated if you added them mid-cycle. Credits never expire but are non-refundable. When the balance hits zero, new calls are blocked until you top up, so production accounts should turn on auto recharge.

The $10 trial credit is granted once per email address, and buying a phone number requires a card on file. Older accounts may still be billed monthly in arrears, and each workspace needs its own payment method.

What does per-minute pricing reward?

A per-minute price makes Retell's revenue a function of how long your calls last, and neutral on whether they end well. A 6-minute call that resolves a booking and a 6-minute call where the caller gives up bill the same. Every per-minute voice platform works this way, and it means the incentive to shorten calls, trim prompts and pick a cheaper model sits entirely with you, and Retell's pricing gives you the levers to do it: the model list, the per-component table and the calculator are unusually transparent.

Bland folds the model into one per-minute rate, so its margin depends on which models it runs for you. Per-resolution pricing, like Zendesk's automated resolutions, pays the vendor only when the AI closes the conversation, which rewards the vendor for claiming resolutions. Per-minute billing sits between the two. You pay for the cost of running the call, whatever the outcome.

For a team that already knows its call volume and can put an engineer on prompt length and model choice, that's a good trade. For a team that wants fewer line items, Bland's per-minute rate (model and voice included, telephony billed separately on its pricing page) or a plan with included minutes is easier to budget.

How much does Retell Enterprise cost?

Retell's pricing page lists Enterprise as "Custom Pricing" for voice and chat, with concurrency starting at 50+ calls and a dedicated server and implementation support as paid add-ons. Its Enterprise page is a sales form with no price. Retell has put a number on it elsewhere, though. Its own guide to what to spend on an AI voice agent, last updated August 11, 2026, says Enterprise plans offer custom pricing for high-volume users, "potentially reducing costs to as low as $0.05 per minute."

That post also quotes stale component prices (Claude 3.5 at up to $0.06 a minute, Bland at a flat $0.09), and it doesn't say which part of the bill the $0.05 covers. In its list, the $0.07 to $0.08 it quotes just above is the voice-engine rate before the model. If the $0.05 replaces the $0.07 voice-engine share of our support stack, the 50,000-minute row falls by $0.02 a minute, or $1,000, to about $4,750 a month. That calculation is ours; Retell hasn't confirmed it.

Third-party guides that rank for this keyword put figures on the entry point. All three come from companies that sell a competing product:

We couldn't confirm the $8,000 on any Retell page live today, but G2's pricing tab loaded for us on a later check the same day and lists an "Enterprise" edition "Starting at $8,000.00" for "White-Glove Service" and "Fully managed agent setup" alongside the Pay As You Go edition at $0.07/minute. It does not mention a $3,000 threshold anywhere on the page, so eesel's $3,000 figure isn't something we could find on G2 today; CloudTalk gives the same threshold without a source. The two $8,000 figures aren't the same claim. Zeeg calls it the Enterprise starting price with no period, and CloudTalk calls it a monthly add-on for managed implementation. If it's a monthly fee, it costs more than the whole 50,000-minute support-stack bill ($5,750), which would wipe out the $1,000 a month that a $0.05 voice-engine rate saves at that volume. So at the volumes on this page, Enterprise is worth asking about for the lower rate; managed setup at $8,000 a month isn't. CloudTalk's $0.05 to $0.07 all-in rate is quoted for 100,000 minutes and up, twice our large row.

Our marketplace search, September 19, 2026:

• AWS Marketplace: no Retell listing. A search for "retell" returned 574 unrelated results, none from Retell.
• Vendr: no Retell page.
• G2 pricing tab: loaded on a later check the same day; confirms Pay As You Go at $0.07/minute and Enterprise "Starting at $8,000.00" (White-Glove Service, fully managed agent setup).
• Capterra: a Retell profile with pricing "Not provided by vendor" and no reviews.

One guide that ranks on page one for this keyword, from 11x, describes HIPAA compliance as a $1,000-a-month add-on. Retell's own plan comparison table ticks HIPAA/BAA on pay as you go and reserves only a custom BAA for Enterprise, so check that claim against the page before you budget for it.

How does Retell compare with Vapi, Bland, ElevenLabs and Synthflow?

The fairest comparison is the same volume on each platform's cheapest purchasable option. At the mid volume, 10,000 minutes a month:

Rival prices are from each vendor's live page on September 19, 2026.

Vapi is the closest match in design: you assemble the same pipeline and pay a platform fee on top of providers at cost. Its calculator puts 10,000 minutes at $819 to $1,290, and it counts telephony on Vapi numbers as free. The catch is concurrency: without a paid package you get 4 simultaneous calls, which a 10,000-minute line will exceed at peaks. Our Vapi guide covers the platform.

Bland prices the model into one rate. Its free Start plan ($0.14 a minute) would be $1,400 at this volume, but it caps you at 100 calls a day, and 2,500 calls over 22 working days is about 114 a day. Build lifts the cap to 2,000 a day for $299 a month and drops the rate to $0.12, which is why it's the row we priced. See our Bland AI guide.

ElevenLabs Agents is the cheapest row here before telephony. Pro includes 1,238 minutes and charges $0.08 for each extra one; the model is added per minute (Gemini 2.5 Flash at $0.0012) and telephony is billed separately. Our ElevenLabs Agents guide goes further.

Synthflow no longer shows a self-serve plan: its pricing page says "Enterprise contracts start at $30,000 annually." That puts it out of reach for the small and mid rows. Our Synthflow guide has the rest.

On price alone, Retell sits in the middle of this group at 10,000 minutes. Against Vapi, its closest match in design, it adds 20 concurrent calls with no package (Vapi gives 4, or 10 with Core; ElevenLabs Pro also includes 20), and it gives the most granular control over which model you pay for. For enterprise voice programs with a contact-center team behind them, Parloa prices by contract, and Voiceflow is the option if your team would rather design flows than wire a pipeline.

What do users say about Retell's pricing?

On G2, Retell AI holds 4.8 out of 5 from 2,638 reviews (checked September 19, 2026). Among the con tags G2 extracts from reviews, "Expensive" appears 87 times, behind "Missing Features" (116) and ahead of "Learning Curve" (80). The pricing complaints are mild and consistent: the bill grows with volume and testing.

• A verified user in Retail (June 26, 2026): "Pricing can also grow depending on call volume."
• A verified user in IT and Services (June 23, 2026): "The per-minute pricing structure can feel a bit steep when you are just starting out and doing a lot of testing."
• A verified user in Real Estate (June 24, 2026): "it is at a point kinda expensive but it is worth it."

None of the three disputes a charge; they describe the per-minute model working as designed.

On Trustpilot, Retell has 4.9 from 814 reviews (checked September 19, 2026). The profile is claimed and on a paid Trustpilot subscription, and Retell invites customers to review; both reviews quoted below are marked "Invited". Two recent reviews mention price, from opposite ends:

• A reviewer in Spain (February 11, 2026): "Very cheap and a good control over everything."
• A reviewer in Canada (February 10, 2026): "It costs more than the other option I was piloting, but it solved all the big problems I had with the other option."

Capterra lists Retell with no reviews yet.

Who is Retell's pricing a good fit for?

Retell's pricing fits teams with phone as a real channel and an engineer who will own the bill: someone who picks the model, keeps prompts under the surcharge threshold, and watches concurrency. Past about 50,000 minutes a month, ask for Enterprise terms before you accept list price.

It's the wrong choice if you want one monthly number with no line items, or if your volume is mostly email and chat tickets rather than calls. For that ticket queue, Macha works differently: it's an AI agent layer on the help desk you already run (Zendesk, Freshdesk, Gorgias, Front, HubSpot or Intercom), billed per ticket, where one thread with one person is one charge however many replies it takes. Plans start at $299 a month for 750 tickets, and setup and monitoring by the Macha team is included; the tiers are on our pricing page. For how voice agents and ticket agents divide the work, see our guide to AI agents for customer service.

How we researched this

We read Retell's pricing page in Chrome and with a scripted browser on September 19, 2026, including the component table's Speech-to-Speech tab, the add-ons, the monthly subscription items and the plan comparison table, and we configured Retell's own calculator (and Vapi's) for the screenshots above; the Retell calculator and chat-rate shots were recaptured the same day to show the add-ons line and the two cheapest chat models. Billing rules come from Retell's documentation on billing, billing exceptions, testing, Conductor, phone numbers, concurrency, LLM options and international calling. We checked AWS Marketplace, Vendr, Capterra, Retell's Enterprise page and blog, and three third-party pricing guides for Enterprise figures, read G2 and Trustpilot reviews in Chrome (an earlier G2 visit hit a bot check, but a later check the same day loaded G2's pricing tab, confirming the $0.07/minute and $8,000 Enterprise figures above), and priced each rival from its own pricing page the same day. We did not build a Retell agent or open an account for this post: Retell's sign-up form asks for a work email address behind a reCAPTCHA, and our rules for test accounts rule out both. Every cost here is list-price arithmetic, not a bill we paid, and we have no usage screen to set against the calculator's $0.109.

Frequently asked questions

How much does Retell AI cost per minute? Retell's voice agents cost $0.07 to $0.31 a minute on pay as you go. The $0.07 floor covers only voice infrastructure and the cheapest voice; with the cheapest model and a Retell US number, a real call starts at $0.0866 a minute. A typical support configuration (GPT 5.4 mini, platform voice, US telephony, knowledge base) is $0.114 a minute.

How much does Retell AI cost per month? There's no monthly plan fee. At list prices, 1,000 minutes costs about $116 a month, 10,000 minutes about $1,144 and 50,000 minutes about $5,750 on our support stack, including phone numbers and extra concurrency. A premium stack with GPT 5.4 and ElevenLabs voices roughly doubles those figures.

Is Retell AI free? Retell gives new accounts $10 of free credit, once per email address, which covers about 88 minutes on our support stack. Testing bills at production rates, and buying a phone number needs a card on file, so a free test ends at web calls and text simulations.

Is there a Retell AI pricing calculator? Yes, on Retell's pricing page. It covers the per-minute components (model, voice, telephony) but not phone numbers, concurrency or long-prompt surcharges, and in our run the add-ons line stayed at $0.000 a minute with all five add-ons shown as selected, so add those by hand.

How much does Retell AI chat cost? Retell bills chat agents per AI message by model. The pricing card says "$0.002+", but the table starts at $0.0005 (GPT 5 nano) and $0.001 (GPT 4.1 nano) and runs to $0.10 (GPT 6 Astra). At 3,000 conversations with 6 AI replies each, that's $9 a month on GPT 5 nano, $135 on GPT 5.4 mini and $450 on GPT 5.4.

How much does Retell AI Enterprise cost? Retell's pricing page says "Custom Pricing". Its own blog says Enterprise can bring costs "as low as $0.05 per minute". Zeeg, which sells a competing product, says Enterprise "starts at $8,000" without a period; CloudTalk, also a competitor, calls about $8,000 a month a managed-implementation add-on and says negotiated rates reach $0.05 to $0.07 a minute all in above 100,000 minutes. Both CloudTalk and eesel put the Enterprise threshold around $3,000 a month of usage. None of these is a quote, so ask sales for one.

Does Retell charge for silence, hold time or transferred calls? Silence and hold time are billed, because speech-to-text keeps listening. After a transfer the AI fee stops and only telephony continues. Calls that fail to connect aren't billed, and voicemail is billed only while the AI is on the line.

Does Retell AI include HIPAA on pay as you go? Retell's plan comparison table ticks HIPAA/BAA on pay as you go and lists a custom BAA, SSO, custom MSA/DPA and role-based access control as Enterprise items. We found no published HIPAA add-on fee on Retell's own site.

What is the cheapest AI calling agent available? Among the five platforms we priced at 10,000 minutes a month, ElevenLabs Agents (about $812 plus telephony) and Vapi ($819 to $1,290) came out lowest, Retell was $1,144, Bland $1,499, and Synthflow starts at $2,500 a month on contract. ElevenLabs' figure excludes telephony and Vapi's assumes 4 concurrent calls, so check both against your peak load.

How good is Retell AI? Retell holds 4.8 out of 5 from 2,638 G2 reviews as of September 19, 2026. Reviewers praise how quickly an agent can be built; the most common complaints are missing features, cost at volume and a learning curve.

Is there a free alternative to Retell AI? None of the rivals we priced is free at production volume. ElevenLabs Agents has a Free plan with 15 minutes a month, Bland's Start plan has no monthly fee and bills $0.14 a minute with a cap of 100 calls a day, and Vapi gives $5 of credit to start. Retell's own $10 credit covers about 88 minutes on our support stack, more than any of those.

Is Retell AI profitable? Retell doesn't publish revenue or profit figures, so there's no sourced answer. What its price list does show is which charge is Retell's own: voice infrastructure at $0.055 a minute, billed on every minute whichever model, voice or carrier you pick. The model, voice and phone line are priced per provider.

What are the alternatives to Retell AI? The closest are Vapi (same assemble-your-own model, platform fee plus providers), Bland (one all-in per-minute rate), ElevenLabs Agents (plans with included minutes) and Synthflow (enterprise contracts from $30,000 a year). Our Retell AI guide compares them on features.

How much do AI voice agents cost in general? On the four self-serve platforms we priced, a support-grade voice minute runs from about $0.08 to $0.15 all in, and a month at 10,000 minutes lands between about $812 (ElevenLabs Agents, before telephony) and $1,499 (Bland Build). A premium Retell stack with GPT 5.4, an ElevenLabs voice, denoising and PII removal is $0.21 a minute, or $2,104 at the same volume. Contract-only platforms start higher: Synthflow's minimum is $30,000 a year.

If your support runs through a help desk rather than a phone line, you can start a Macha trial with $50 of free usage and no card, or compare our tiers on the pricing page.

Sources: Retell AI pricing · Retell billing · Retell billing exceptions · Retell testing pricing · Retell concurrency and limits · Retell LLM options · Retell international calling · Retell phone numbers · Retell Conductor billing · Retell AI QA · Retell: How much should you spend on an AI voice agent? · Retell Enterprise plan · G2: Retell AI reviews · G2: Retell AI pricing · Trustpilot: Retell AI · Capterra: Retell AI · Zeeg: Retell AI pricing guide · CloudTalk: Retell AI pricing · eesel AI: Retell AI pricing · Vapi pricing · Vendr: Vapi · Bland pricing · ElevenLabs Agents pricing · Synthflow pricing · 11x: Retell AI pricing · Zendesk: outcome-based pricing