Retell AI Pricing Explained: What You Actually Pay Per Minute

Retell AI pricing is usage based: you pay for the minutes your AI voice agents spend on calls, with no monthly platform fee and $10 in free credits to start. The published envelope runs from $0.07 to $0.31 per minute, and where you land inside it depends on three choices: which voice you use, which language model runs the conversation, and how you handle telephony.

A 10,000-minute month lands around $900 at the bottom of that range and over $3,000 at the top, so the three configuration choices matter more than the headline number. This guide breaks down each component with Retell's own published numbers, works through what typical setups cost per minute and per month, and covers the add-on fees that surprise teams after launch. All prices below come from Retell AI's pricing page and billing documentation, verified on July 20, 2026. Retell can change them at any time.

How Retell AI pricing works#

Retell AI is a developer platform for building custom voice agents, and its pricing mirrors that architecture. Every call minute is assembled from three metered components:

1. The voice engine. Retell's conversation infrastructure plus text-to-speech. This is the $0.07 per minute base most coverage quotes: $0.055 for voice infrastructure and $0.015 for platform text-to-speech. Premium voices from ElevenLabs or OpenAI push it to roughly $0.08.
2. The language model. The LLM that decides what the agent says, billed per minute at rates that vary by model.
3. Telephony. Carrying the actual phone call, at $0.015 per minute for US calls on Retell's managed Twilio, plus $2 per month for each phone number.

There is no seat license and no required subscription. You deposit credits, calls draw them down, and billing is prorated to the second after any per-call minimums.

Retell AI pricing at a glance#

Retell AI does not sell tiered subscriptions. It sells metered usage on top of a free credit grant, with a custom enterprise track above it. Here is the whole structure in one view before the component math.

The rest of this guide unpacks the per-minute rate, because that is where the pay-as-you-go number actually gets decided.

The per-minute cost of each component#

Voice engine: $0.07 to $0.08 per minute#

The floor for any Retell call. Retell's platform voices bill at $0.07 per minute all-in (infrastructure plus text-to-speech); ElevenLabs and OpenAI voices raise it about a cent.

Language model: $0.003 to $0.16 per minute#

The biggest swing in the whole stack, a 50x spread depending on model choice:

Cheap models handle simple, scripted interactions. Conversations that need objection handling generally push teams toward the $0.045 to $0.08 tiers, which is where per-minute costs double.

Telephony: $0.015 per minute plus numbers#

US calls on managed Twilio cost $0.015 per minute, and each phone number runs $2 per month. A verified phone number (branded caller identity) is $10 per month. You can bring your own Twilio account instead, in which case you pay Twilio directly.

What realistic configurations cost#

Stacking the components gives you the per-minute rate that lands on the invoice. Four representative builds:

Add-ons (below) are how a frontier build reaches the $0.31 top of Retell's published range.

Translated to monthly bills at the Standard build:

• 1,000 minutes: about $130, plus $2 per number.
• 5,000 minutes: about $650.
• 20,000 minutes: about $2,600, before any volume negotiation.

A typical 4-minute qualification call at the Standard build costs roughly $0.52. That is the number to compare against what a missed or slow-answered lead costs you; our cost breakdown of automated appointment setting versus a human ISA runs that comparison in detail.

The add-on fees that stack up#

Retell's base math is transparent, but production deployments usually turn on features that bill separately:

• Knowledge base: $0.005 per minute when the agent uses it, and $8 per month per document beyond the first 10.
• Advanced denoising: $0.005 per minute.
• PII removal: $0.01 per minute.
• AI quality assurance: $0.10 per minute after the first 100 free minutes. Turned on across all traffic, this single feature nearly doubles a Standard build.
• Branded calls: $0.10 per outbound call.
• Extra concurrency: 20 simultaneous calls are included; each additional slot is $8 per month.

Two billing rules in Retell's documentation are worth knowing before you dial at volume: calls that use a dynamic opening message bill a 10-second minimum even if the recipient hangs up instantly, and unusually long prompts incur extra token charges. On large outbound lists where most dials go unanswered or end in seconds, minimums like these shape your cost per connect.

If AI quality assurance runs on every call, stop budgeting the $0.13 Standard rate: the $0.10 per minute add-on nearly doubles it, and that higher number is the one to plan around.

What is included without extra charges#

• $10 in free starting credits, full platform access, no monthly platform fee
• 20 concurrent calls
• First 100 minutes of AI quality assurance
• First 10 knowledge base documents
• Call recordings, transcripts, and the agent-building tooling itself

Enterprise plans add custom pricing, dedicated infrastructure, uncapped concurrency, and 24/7 support with an account manager. Retell does not publish enterprise rates; volume discounts are negotiated.

Is Retell AI pricing worth it?#

For the buyer Retell is built for, yes. It is one of the most transparent pricing models in the voice AI category: every component is published, there is no forced subscription, and $10 of credits buys a real test. In our comparison of AI cold calling software, Retell held the middle path between raw infrastructure and finished product. If you are weighing it against the rawest option on that spectrum, our Vapi pricing breakdown does the same component math for the $0.05 platform.

What the meter leaves out is the work. Retell sells the platform and leaves the caller to you: conversation design, prompt engineering, integrations, and ongoing tuning are your job, which in practice means developer time on top of every per-minute rate in this guide. A $0.13 per minute build with an engineer spending ten hours a month tuning it is not a $0.13 per minute system.

Budget for both lines and weigh them against what the calls produce. A qualification call that books an appointment stays cheap even at five times these rates. A poorly designed agent that burns through a lead list is the expensive outcome, whatever the meter reads.

Who Retell AI fits (and who it does not)#

The pricing model tells you who Retell is built for. Every rate in this guide is a lever you set yourself, which is the whole appeal for teams that want to tune their own stack. Teams that just need calls placed inherit a build project instead.

Retell AI fits you if:

• You have a developer, or budget to hire one. Conversation design, prompts, integrations, and tuning are your work, and they sit on top of every per-minute rate above.
• You like per-component cost control. Choosing a cheaper model or skipping an add-on lowers the bill directly, and the $10 credit grant plus no platform fee lets you validate a build before you commit spend.
• You run regulated outbound and can build to it. Retell states HIPAA, SOC 2 Type II, and GDPR compliance, and PII redaction is available as a $0.01 per minute add-on.
• You expect serious volume, where enterprise pricing and uncapped concurrency start to matter more than the per-minute rate.

Retell AI is a harder fit if:

• You need calls going out this week and have no engineer to point at it. The meter does not include the build, and the build is the slow part.
• You cannot carry developer time on top of the rate. That $0.13 Standard build still needs someone tuning it every month, and that person's time is the cost the per-minute number hides.
• You expect objection handling to work on day one. Out of the box that follows the design work you put in, and a higher rate does not buy it.

Where Rezora IO fits#

Start with who should skip Rezora IO. If you have a developer who wants API-level control over the model, the prompts, and the call flow, buy Retell and enjoy it. That is exactly what Retell is built for, and a done-for-you caller would only take that control away. Rezora IO is for the other reader: the operator who wants the calls made and has no interest in building anything to get there.

What lets it skip the build is the model underneath. Rezora IO trains its own hosted caller on sales conversations instead of handing you a foundation model to prompt, so objection handling ships in the box. You connect your contacts or upload a CSV, and it starts dialing new leads, qualifying them, and booking callbacks with nothing to configure first.

The decision is not really about price, since both products bill by usage. It comes down to which job you are hiring for: a platform to build a voice agent on, or a caller that already works. Pricing is usage based, with enterprise plans and no per-component meter to assemble.

The two products line up like this on the decisions that drive cost and effort:

FAQ#

How much does Retell AI cost per minute?#

Between $0.07 and $0.31 per minute, per Retell's published pricing. The voice engine costs $0.07 to $0.08, the language model adds $0.003 to $0.16 depending on the model, and US telephony adds $0.015. Most production configurations land between $0.09 and $0.26 before optional add-ons.

Does Retell AI have a monthly fee?#

No platform fee or required subscription. You pay per minute of usage, plus small monthly line items if you use them: $2 per phone number, $8 per extra concurrency slot beyond the included 20, $8 per knowledge base document beyond 10, and $10 for a verified number.

Is there a Retell AI free trial?#

New accounts get $10 in free credits with full platform access, which covers roughly an hour of calling at mid-range configurations. There is no time-limited trial to expire; credits simply draw down.

What does Retell AI charge for concurrent calls?#

Twenty simultaneous calls are included on standard accounts. Additional concurrency costs $8 per slot per month, and enterprise plans remove the cap.

What is not included in the per-minute rate?#

Optional features bill on top: knowledge base usage, denoising, PII removal, AI quality assurance at $0.10 per minute after the first 100 minutes, and branded outbound calls at $0.10 per call. Watch the billing exceptions too: dynamic opening messages carry a 10-second minimum per call, and very long prompts incur token surcharges.

Is Retell AI HIPAA compliant?#

Retell AI states that it is compliant with HIPAA, SOC 2 Type II, and GDPR, and PII removal is available as an add-on at $0.01 per minute. In practice a stated certification is not a signed agreement, so ask for the BAA and confirm the current terms with Retell before you run protected health information through it.

Is Retell AI worth it?#

For teams that want to build and control their own voice agent, yes: the pricing is transparent, there is no forced subscription, and $10 in credits is enough to test a build before you commit. The cost the meter hides is developer time for prompt design, integrations, and tuning, which is ongoing and easy to underestimate. If you want calls made instead of a platform to build, a trained caller answers a different question.

How does Retell AI compare to other voice AI platforms?#

Retell sits in the middle of the market: more finished than raw infrastructure, less finished than a done-for-you caller. Our comparison of AI cold calling software places it against the field, and our Vapi pricing breakdown runs the same component math on the rawer end. If you want a caller that ships pre-trained instead of a platform to assemble, that is where Rezora IO fits.

Do you need a developer to use Retell AI?#

Retell offers both a no-code builder and a full SDK, so a simple agent can be assembled without writing code. Production deployments usually need engineering anyway: prompt design, integrations, and tuning are what turn a working demo into a reliable caller. Budget for that time on top of the per-minute rates, whichever path you take.