Retell AI Pricing: Retell Pricing per Minute, Retell AI Cost and Monthly Bills at Real Call Volumes

Retell AI costs nothing to start and has no monthly platform fee. You pay per connected minute, tracked to the second, and a new account gets $10 of free credit. Retell describes it as "true pay as you go," and the price list backs that up: there is no minimum, no contract and no feature gating between the pay as you go and Enterprise plans. What you pay for is four meters that run on every minute of every call.

The advertised range is "$0.07 to $0.31 / min for AI Voice Agents." The low end is simply infrastructure plus a $0.015 voice plus the cheapest model: $0.055 plus $0.015 plus $0.0016 comes to $0.0716. It does not include telephony, so on a Retell number the true floor is about $0.087. The high end is not a ceiling at all. The rate card lists GPT 6 Astra at $0.32 a minute for the model alone, which puts a call on that model at roughly $0.40 a minute with a platform voice and US telephony, and about $0.72 on the fast tier.

The model is the only meter that moves by a factor of 100, so it decides most of your bill. These are the standard tier rates from Retell's rate table, with the all-in per-minute cost on a Retell platform voice and a US Twilio number ($0.055 plus $0.015 plus the model plus $0.015).

Swap in an ElevenLabs voice and every row goes up by $0.025. Retell also lists speech-to-speech models separately: GPT Realtime and GPT Realtime 1.5 at $0.345 a minute, GPT Realtime 2 and 2.1 at $0.38, and the Realtime mini models at $0.07.

Retell marks GPT 5.6 Terra, Claude 5 Sonnet, GPT 5.5 and GPT 5.4 as recommended. That is a useful signal about quality, and an expensive one if you follow it blindly: GPT 5.4 costs 3.3 times GPT 5.4 mini per model minute. For appointment booking, order status and after-hours message taking, the small models are usually enough. Reserve the large ones for calls that reason through several steps, like insurance intake or a complex reschedule.

If you priced Retell with the estimator on its pricing page, check your number against the rate table on the same page. On September 18, 2026 the two did not match. The estimator's code carries its own list of model rates, and for several smaller models they are higher than the published table.

The practical effect is small per minute and real per quarter. A 20,000 minute line on Gemini 3.0 Flash comes out about $220 a month higher in the estimator than on the rate card, which makes Retell look more expensive than it is on budget models. A line using PII removal comes out $100 a month lower. The rate table is the more specific document, so use it for a budget and treat the calculator as a rough first pass. Your invoice is the final word: ask Retell which rate applies to your model before you commit a volume forecast to finance.

Monthly cost on pay as you go, with a Retell platform voice, a US Twilio number and no add-ons, at four model choices:

Add $2 a month per Retell phone number. If you bring your own SIP trunk or telephony, subtract $0.015 a minute, which is $150 at 10,000 minutes, and pay your carrier instead.

How many minutes will a US business actually use?

A dental or medical front desk taking 50 calls a day at three minutes each uses about 4,500 minutes a month over 30 days. A home services company running inbound booking plus reminder calls at 150 calls a day and two minutes each uses about 9,000. An outbound qualification campaign is harder to size, because Retell bills only connected calls but charges for time spent on voicemail while the agent is active. Run a pilot week before you forecast.

Add-ons that change the per-minute rate

Retell prices extras per minute or per call: knowledge base $0.005 a minute, advanced denoising $0.005, safety guardrails $0.005, PII removal $0.01, batch calling $0.005 per dial, branded caller ID $0.10 per outbound call, and AI quality assurance $0.10 a minute after the first 100 free minutes. A support line with a knowledge base, denoising and PII removal adds $0.02 a minute, or $200 at 10,000 minutes. Quality assurance is the one to watch: switched on for every call it costs more than the rest of the minute combined.

Concurrency is where Retell is most generous. Every account includes 20 simultaneous calls at no charge, and extra capacity costs $8 per concurrent call per month, added or removed at any time. Enterprise plans offer custom concurrency "starting at 50+ calls."

Size concurrency from your busiest hour, not your monthly total. A line that takes 120 three-minute calls between 8 and 9 a.m. needs roughly 6 lines on average in that hour and more at the peak minute, which still fits inside the free 20. A 50 line outbound dialer pays for 30 extra lines, $240 a month, which on a 50,000 minute month adds about half a cent per minute.

For comparison, Vapi includes 4 concurrent calls and charges $10 a month for each extra line, so the same 20 lines would cost $160 a month there. Bland caps concurrency by plan, at 10 on Start and 50 on Build.

Four billing rules from Retell's pricing FAQ decide whether your invoice matches your forecast.

Is Retell billed per second or rounded up per call?

Per second. Retell says each call "is tracked to the nearest second" and charged on total accumulated minutes at the end of the billing cycle, with "no rounding up per call." Short calls are cheap, which matters for outbound campaigns full of 20 second calls.

Am I charged during silence or hold time?

Yes. The whole call is billed "because the speech-to-text engine remains active and listening throughout." An agent that puts a caller on hold while it looks up an order costs the same per minute as one that is talking, so keep lookups fast and hold music short.

Does billing continue after a transfer to a human?

Only the telephony part. Once the call transfers, "the AI voice agent fee stops," and the telephony fee continues for the rest of the call. On a Retell number that is $0.015 a minute, so a 10 minute conversation with a human after a transfer costs 15 cents.

Do I get charged for failed calls or voicemails?

Failed calls are not billed. Calls that reach voicemail are billed "only for the duration the AI agent is active on the line," so an agent that leaves a 30 second message pays for 30 seconds.

The three most-compared voice agent platforms price the same minute in different ways. Vapi and Retell charge a platform fee and pass through models; Bland charges one rate with the model and voice included. At 10,000 US minutes a month, using each vendor's own published rates:

On a small model Vapi's per-minute bill is slightly lower than Retell's, mostly because Vapi numbers carry no telephony charge. Retell pulls ahead when you need concurrency without buying a package, and when you want every model's price on one published table. Bland is the choice when finance wants a single rate that never moves with model prices. We break each alternative down at three volumes in our guide to Vapi alternatives and Retell compared per minute, the full Vapi bill is on our Vapi pricing page, and Bland's plan-by-plan math is in Bland AI pricing.

Is Retell AI cheaper than Bland?

On most model choices, yes. Retell on GPT 5.4 mini costs about $0.109 a minute all in on a US line, against Bland's $0.14 on Start or $0.12 plus $299 a month on Build, before Bland's telephony. Bland only comes out ahead once you pick a large model on Retell: GPT 5.4 puts Retell at $0.165 a minute.

Retell publishes no Enterprise price. The plan is quoted by sales and adds, in Retell's list, "volume pricing that scales as you grow," a dedicated stable server as an add-on, custom MSA and DPA terms, role-based access control, custom SSO, a high cap on concurrent calls and 24/7 support through a dedicated portal, with dedicated implementation support and a named account manager.

For US healthcare buyers, Retell describes itself as "SOC 2 certified, HIPAA-ready," and lists HIPAA with a BAA and a custom BAA in its compliance comparison. Confirm in writing whether the BAA you need is available on pay as you go or only under an Enterprise agreement before you route patient calls through it, and ask for the data retention period on recordings and transcripts, which Enterprise lets you customize.

Before signing an Enterprise contract, get three numbers in writing: the effective per-minute rate at your committed volume for the model you will actually run, the included concurrency, and whether model rates are locked for the term. Retell's model list changes often as new models are added and older ones are repriced, so a rate lock is worth more than a small headline discount.

Retell is worth it for teams that want a voice agent they configure themselves, with transparent per-component pricing, 20 free concurrent lines and no package to buy. At about $0.11 a minute on a capable small model, a 5,000 minute front desk line costs roughly $545 a month, less than a day of a US receptionist's time.

The minutes are not the whole cost, though. Retell is a build platform: someone has to write prompts, connect the calendar or CRM, test edge cases and review calls. For a business without that person, the honest total includes a solution partner or an internal owner, and that usually dwarfs the minute bill.

It is also worth checking how many of those calls need a phone agent at all. When a large share are order status questions, refund requests or the same support questions, deflecting them to email and chat stops them before they cost a minute. Our catalog lists ready-made agents for exactly that work, like a support agent that answers from your help docs, an order concierge and a refund resolver, each with a flat monthly price and its permissions shown before you deploy. Start with AI agents for customer service, or compare the outcome-priced route with Fin's per-resolution pricing and Sierra AI pricing.