Retell AI Pricing 2026: Per-Minute Cost of Voice, LLM and Telephony
Short answer: Retell AI has no subscription. A voice agent costs $0.07 to $0.31 a minute on pay-as-you-go, built from four parts: Retell's voice infrastructure at $0.055 a minute, a voice from $0.015, a language model from a fraction of a cent to $0.64, and US telephony at $0.015. New accounts get $10 in free credit, and Enterprise pricing is custom. Verified on Retell's pricing page, October 2026.
The range is wide because you choose every component. A lean receptionist can run under 9 cents a minute; the same agent with a premium voice, a top model and a toll-free line costs three times that. Below are three typical builds priced per minute and at 10,000 minutes a month, then the billing rules that push real invoices above the calculator.
How Retell AI pricing is built
| Component | What it covers | Price per minute | Source |
|---|---|---|---|
| Retell Voice Infra | The conversation voice engine, charged on every call | $0.055 | Retell pricing |
| Voice (text-to-speech) | Retell Platform, Cartesia, OpenAI, Minimax, Fish or Inworld voices | $0.015 | Retell pricing |
| Voice, ElevenLabs | Flash, Multilingual v2, v3, v4 and Turbo models | $0.040 to $0.10 | Retell pricing |
| Language model, Standard | 30+ models from OpenAI, Anthropic and Google | $0.0016 to $0.32 | Retell pricing |
| Language model, Fast Tier | Priority processing on supported models | $0.0048 to $0.64 | Retell pricing |
| Telephony, US | Retell numbers on Twilio | $0.015 | Retell pricing |
| Telephony, your own | SIP trunking or custom telephony | No Retell charge | Retell pricing |
Component prices read on Retell's pricing page on October 30, 2026. Fixed monthly items and add-ons are covered further down.
Dupple data: Toolradar, our software directory, ranks 6 AI voice agent platforms as of October 30, 2026; 3 of them have a free plan, and the median cheapest paid plan across those that publish one is $179.50 a month (Toolradar's AI voice agent ranking). Retell has no plan fee at all, so a low-volume team pays only for the minutes it uses.
What does a Retell AI voice agent cost per minute?
Here are three builds a small business might actually run, each priced from Retell's list. All totals are our arithmetic, before any per-number or concurrency fees.
| Build | Voice | Model | Telephony | Add-ons | Per minute | 10,000 minutes |
|---|---|---|---|---|---|---|
| Lean receptionist | Retell Platform voice ($0.015) | Claude 5.5 Haiku ($0.0032) | US Twilio ($0.015) | None | $0.0882 | $882 |
| Retell-recommended default | Retell Platform voice ($0.015) | GPT 5.6 Terra ($0.064) | US Twilio ($0.015) | None | $0.149 | $1,490 |
| Premium support line | ElevenLabs v3 Conversational ($0.08) | Claude 5.5 Sonnet ($0.064) | US Twilio toll-free ($0.06) | Knowledge base ($0.005), PII removal ($0.01) | $0.274 | $2,740 |
Each row adds the $0.055 voice infrastructure fee. Rates from Retell's pricing page, October 30, 2026.
The model choice is the biggest lever. Moving the lean build from Claude 5.5 Haiku to Claude 5.5 Sonnet adds $608 per 10,000 minutes, more than the $550 that same volume pays in infrastructure fees. Start with the cheapest model that handles your call script reliably, and move up only for the calls that need reasoning.
The voice is the second lever. ElevenLabs Flash costs $0.025 a minute more than a platform voice, or $250 per 10,000 minutes; ElevenLabs v3 and v4 cost up to $0.085 more. For a receptionist that books appointments in a calendar, a $0.015 voice is often enough. For a brand-facing sales line, the premium voice may be worth the money.
The floor on Retell's "from $0.07" headline is easy to reproduce: voice infrastructure, a platform voice and GPT 5 nano at $0.0016, on your own telephony, comes to $0.0716 a minute (our arithmetic). Few production agents run that lean, because a nano-class model struggles with anything beyond a short script.
Fast Tier and speech-to-speech models
Retell lists a second LLM column called Fast Tier, which buys more predictable latency. On this price list it doubles the model rate: GPT 5.4 goes from $0.080 to $0.16 a minute, and GPT 5.5 from $0.16 to $0.32. Use it for live sales calls where a pause costs the deal, not for after-hours message taking.
Retell also lists speech-to-speech models, which handle listening and speaking in one model. GPT Realtime mini and GPT Realtime 2.1 mini are listed at $0.07 a minute, while GPT Realtime 2 and 2.1 are $0.38. Retell's page lists these rates separately from the component stack, so compare them against a full build total, not against the model line alone.
Monthly fees and add-ons
Per-minute costs are most of the bill, but these items land on the invoice too, per Retell's pricing page, October 30, 2026:
| Item | Price | What it means |
|---|---|---|
| Retell phone number | $2 a month | No charge if you bring your own numbers |
| Concurrent calls | First 20 free, then $8 per call per month | A 30-line contact center pays $80 a month extra (our arithmetic) |
| Knowledge bases | First 10 free, then $8 each per month, prorated | Plus $0.005 a minute when an agent uses one |
| Retell SMS | $20 a month | Messages cost $0.01 each |
| Branded caller ID | $0.10 per outbound call | Adds up fast on outbound campaigns |
| Batch calling | $0.005 per dial | Charged on outbound lists |
| Advanced denoising, safety guardrails | $0.005 a minute each | Optional per agent |
| PII removal | $0.01 a minute | Redacts personal data from transcripts and recordings |
| AI Quality Assurance | First 100 minutes free, then $0.10 a minute | Automated call scoring |
| Verified phone number | $10 | Retell's page shows it as both a monthly and a one-time fee; confirm in your dashboard |
For an outbound team, branded caller ID is the add-on to model. At $0.10 per call, 5,000 outbound dials a month adds $500 before anyone picks up (our arithmetic), more than the per-minute cost of many short calls.
Billing rules that raise the real cost
Retell bills calls to the nearest second and sums them at the end of each cycle, with no per-call rounding. Several rules in its FAQ and billing exceptions documentation still push invoices above the calculator:
- Long prompts scale the billed duration. When an agent's prompt passes 4,000 LLM tokens, Retell multiplies the billed duration by prompt tokens divided by 4,000. Retell's own example: a 60-second call with 4,800 tokens bills as 72 seconds. Tool descriptions, the transcript so far, tool results and retrieved knowledge base content all count toward the total, so long calls drift past the threshold.
- Silence and hold time are billed. The speech engine keeps listening for the whole call, so a caller on hold costs the same per minute as a caller talking.
- A 10-second minimum applies to calls that open with a dynamic AI first sentence. A static first sentence avoids it.
- Voicemail is billed for the time the agent is active on the line. Calls that fail to connect are free.
- Transfers stop the agent fee but not telephony. After a warm or cold transfer to a person, only the telephony rate continues.
- Burst concurrency costs extra. Retell's changelog describes a burst option above your concurrency limit, billed at $0.1 a minute for burst usage.
The token rule has already moved once. Retell's May 2025 changelog introduced the scaling at 3,500 tokens; the current documentation sets the threshold at 4,000. Keep prompts tight and use knowledge base retrieval for reference material, rather than pasting everything into the system prompt.
Is Retell AI free?
There is no free plan, but there is no fee to start. New accounts get $10 in free credit with full platform access, which covers about 113 minutes on the lean build above (our arithmetic). That is enough to build and test an agent, not to run a pilot with real callers.
Chat agents are billed separately, from $0.002 per message, per Retell's pricing page. A team that wants both a phone agent and a website chat agent pays for each channel's usage.
Pay-as-you-go vs Enterprise
| Pay-as-you-go | Enterprise | |
|---|---|---|
| Price | $0.07 to $0.31 a minute | Custom pricing |
| Concurrent calls | 20 included, more on demand | No cap; custom concurrency from 50+ calls |
| Server | Shared | Dedicated server available as an add-on |
| Support | Community and email | 24/7 support portal, implementation help, named account manager |
| Contracts | None | Custom MSA, DPA and BAA, SSO, role-based access |
Plan comparison from Retell's pricing page, October 30, 2026.
Most teams under 20 simultaneous calls never need Enterprise. The trigger is usually compliance paperwork (a custom BAA or DPA), SSO, or a dedicated server, not price. Retell's page describes the platform as SOC 2 certified and HIPAA-ready.
Is Retell AI cheaper than building on Vapi?
It depends on the build, so compare like for like. Retell charges its fixed infrastructure fee on every minute and sells voices and models from one price list; our Vapi vs Retell AI cost breakdown prices the same support line on both platforms. For the wider field, see the best AI voice agents; for voice options beyond Retell's list, see AI voice generators and AI text-to-speech tools. To clone or design a custom voice first, read how to create an AI voice. Toolradar keeps a Retell AI pricing profile with plan details.
FAQ
How much does Retell AI cost per minute in 2026?
Retell AI costs $0.07 to $0.31 a minute on pay-as-you-go, as of October 2026. Every minute includes $0.055 for voice infrastructure, plus a voice, a language model and telephony. A Retell-recommended default build with a platform voice and US telephony comes to $0.149 a minute (our arithmetic).
Does Retell AI have a monthly fee?
No plan fee. You pay per minute, plus small monthly items if you use them: $2 per Retell phone number, $8 per concurrent call beyond the first 20, and $8 per knowledge base beyond the first 10.
How much does 10,000 minutes on Retell AI cost?
Between about $882 and $2,740 a month for the three builds on this page, depending mainly on the model and voice (our arithmetic). The infrastructure share is fixed at $550 per 10,000 minutes; everything else depends on what you choose.
Does Retell charge for silence or failed calls?
Silence and hold time are billed, because the speech engine keeps listening for the whole call. Calls that fail to connect are free, and voicemail is billed only while the agent is active.
Can I use my own phone numbers with Retell?
Yes. Retell charges nothing for SIP trunking or custom telephony, and custom numbers carry no Retell number fee. Your carrier bills you directly instead of the $0.015 a minute Retell charges for US Twilio calls.
Is Retell AI HIPAA compliant?
Retell's page describes the platform as SOC 2 certified and HIPAA-ready and lists HIPAA and BAA in its plan comparison. Custom BAA, MSA and DPA terms are part of Enterprise, so regulated teams should plan for a sales conversation.
How we checked these prices
We read Retell's pricing page, its billing exceptions documentation and its changelog on October 30, 2026. The three builds use US telephony and 10,000 minutes a month; every per-minute total and monthly figure is our arithmetic on printed rates. Retell changes model and voice prices often, so check the calculator on its pricing page before you budget.
Cite this: Dupple, "Retell AI Pricing 2026: Per-Minute Cost of Voice, LLM and Telephony", November 2026.
A voice agent's price is a stack of choices. Want more cost breakdowns like this for AI at work? Try Dupple X →
Related
- Vapi vs Retell AI: the same line priced on both platforms.
- Best AI voice agents: Retell against other platforms.
- Best AI voice generators: voices beyond Retell's list.
- Retell AI on Toolradar: the listing with plan details.
- Dupple learning hub: more guides on AI tools and costs.
Some offers on this page may be paid placements or contain affiliate links.
Across our whole catalogue the median is $23.50 a month and 37% of tools advertise a free tier. Figures are the lowest recurring monthly price quoted in our own review of each tool, read on 2026-10-23. Full method and every category.
