Retell AI Pricing 2026: True Per-Minute Cost vs Marketing Claims
TL;DR: Retell AI Pricing in 60 Seconds
Retell AI advertises a $0.07 to $0.31 per-minute range. The honest middle of the curve sits around $0.10 to $0.18 per minute when you add LLM, TTS, and voice infrastructure costs together. On top of per-minute rates you pay $2 per phone number, $8 per concurrent call slot beyond the included 20, and $8 per knowledge base beyond the first 10. There is no platform fee on the pay-as-you-go tier. Enterprise unlocks dedicated infrastructure and HIPAA, with custom pricing.
Want a Managed Voice AI Instead of a Toolkit?
AINORA is a managed end-to-end voice AI service rather than a developer platform. Hear it on a real phone call: +1 (518) 241-8125 (US) or +370 5 200 2620 (LT). No signup, no API keys, no concurrency math.
Retell AI is one of the most popular developer-facing voice AI platforms in 2026. The pricing page is more transparent than most, but it still hides the real cost behind a multi-component breakdown. This post unpacks what you actually pay per minute, what extras land on the invoice, how Retell compares with Vapi, ElevenLabs Conversational AI, and Bland, and when a managed voice AI service makes more sense than a DIY developer platform.
Per-Minute Breakdown by Component
Retell prices voice agents as a stack of three components: the LLM, the voice infrastructure, and the text-to-speech engine. The total per-minute cost is the sum of whatever you choose at each layer.
- LLM cost: $0.003 to $0.080 per minute, depending on which model you select. Smaller and faster models sit at the low end. Frontier reasoning models sit at the top.
- Voice infrastructure: $0.055 per minute, flat. This covers Retell's telephony plumbing, audio handling, and conversation orchestration.
- Text-to-speech: $0.015 to $0.040 per minute. Retell platform voices, Minimax, Fish, Cartesia, and OpenAI voices sit at $0.015. Premium voices via the third-party voice provider with the highest brand recognition sit at $0.040.
Add those together and a realistic mid-tier configuration (good LLM, premium TTS) lands around $0.10 to $0.18 per minute before any extras. The $0.07 floor only applies if you pick the cheapest LLM and the cheapest TTS, which is a real choice but not the configuration most production deployments end up at.
Voice Options and Their Real Cost
Retell supports multiple TTS providers natively. The cost differential matters more than buyers usually expect, because TTS is the layer the consumer actually hears.
- Retell platform voices, Minimax, Fish, Cartesia, OpenAI voices: $0.015 per minute each.
- Premium third-party voices: $0.040 per minute. Higher fidelity, more languages, but 2.7x the TTS cost.
For chat agents (text-only), Retell charges from $0.002 per message. That is a different product and usually only relevant if you are building a multi-channel agent that spans voice and chat.
Integrations Included
Retell ships native integrations with Make, Twilio, Vonage, Go High Level, n8n, and HubSpot. That covers the most common workflow tools. Anything outside that list is your own engineering work, which is fine for a developer platform but is worth budgeting for honestly. Most production teams spend more time on integration plumbing than on the voice agent itself.
Hidden Costs Most Buyers Miss
Per-minute is the headline number. The line items below are the ones that show up on month two of the invoice and surprise procurement.
- Phone numbers: $2 per number per month. Trivial at small scale, meaningful at scale (100 DIDs is $200 per month before a single call).
- Concurrent call capacity: 20 included on pay-as-you-go. Each extra concurrency slot costs $8 per month. A campaign that needs 100 concurrent calls is $640 per month in concurrency before any per-minute usage.
- Knowledge bases: First 10 free, then $8 per knowledge base per month. Multi-tenant deployments hit this fast.
- Verified phone numbers: $10 per verified number per month. Required if you want clean caller-ID display on outbound campaigns.
- Add-on features: Branded calls, PII removal, and other premium features carry surcharges. Not all of them are listed at the same level of detail as the per-minute rate.
True Cost vs Marketing Claim
Do This Math Before You Sign
Take your monthly minutes, multiply by your realistic per-minute cost (use $0.14 as a midpoint estimate), then add concurrency slots, phone numbers, knowledge bases, and any verified-DID fees. That is your real monthly bill on Retell. Compare that against Bland's flat $0.11 to $0.14 plus platform fee, and against managed alternatives where the per-minute number includes everything.
Worked example: a sales team running 50,000 outbound minutes per month with 100 concurrency, 10 phone numbers, and 15 knowledge bases. On Retell at $0.14 average per minute, that is $7,000 in usage, $640 in concurrency overage, $20 in DIDs, and $40 in knowledge bases. Total: $7,700. Same volume on Bland Scale at $0.11 plus $499 platform: $5,500 + $499 = $5,999. The headline rates do not tell you which is cheaper. The full math does.
Retell vs Vapi vs ElevenLabs Conversational AI vs Bland
| Feature | Retell AI | Vapi | ElevenLabs Conv. AI | Bland |
|---|---|---|---|---|
| Pricing Model | Per-component per-minute | Per-component per-minute | Credit-based per-minute | Flat per-minute + platform fee |
| Advertised Floor | $0.07/min | Component-stack (typically $0.05-$0.30+) | Tied to credit plan | $0.11-$0.14/min |
| Realistic Mid-Tier | $0.10-$0.18/min | $0.23-$0.33/min (per Bland comparison) | Variable by voice | $0.11-$0.14/min flat |
| Voice Quality Tier | Multiple TTS providers, 100s of voices | Multiple TTS providers | 10,000+ voices, voice cloning | Native voice plus options |
| Concurrency Included | 20 (pay-as-you-go) | Variable | Plan-dependent | 10 (Start) to 100 (Scale) |
| Platform Fee | None on PAYG | None advertised | Credit subscription | $0 / $299 / $499 / Custom |
| Phone Number Fee | $2/month | Yes | Yes (via Twilio) | Included |
| Compliance | HIPAA on Enterprise | Enterprise tier | SOC 2, HIPAA, GDPR | Enterprise tier |
| Best For | Developers wanting flexibility | Developers wanting flexibility | Brands prioritising voice quality | Teams wanting flat-rate predictability |
DIY Platforms vs Managed Voice AI
Retell, Vapi, ElevenLabs Conversational AI, and Bland are developer platforms. They give you the building blocks - LLM, TTS, telephony, function calling - and you assemble the agent. That is the right model if you have engineering capacity, want full control over the prompt, the tooling, the integrations, and the data flow, and are happy owning the operational burden of voice infrastructure (recordings, transcripts, retries, compliance, monitoring).
Managed voice AI is a different product entirely. Instead of an API and a credit balance, you get a working agent on a working phone number, with the prompt, the tools, the database, the recordings, the transcripts, the compliance posture, and the monitoring dashboard already in place. You hand over the use case and the script, and the vendor ships a tested agent. The trade-off is less low-level control in exchange for time-to-value measured in days rather than months.
AINORA sits in the managed category. The dental, debt collection, restaurant, and legal demos are all live phone numbers you can call right now. Pricing is per deployment with per-minute usage rolled in, not a multi-line-item dance with concurrency slots and verified-DID fees. That is not better or worse than a developer platform. It is a different buyer choice.
Who Should Choose Each
- Choose Retell AI if you have an engineering team, want fine-grained control over the LLM and TTS stack, and are building a custom voice product where the agent itself is part of your IP.
- Choose Vapi if you prefer maximum component flexibility and are comfortable with the per-component pricing model.
- Choose ElevenLabs Conversational AI if voice quality is the single most important variable - for example, brand voice agents, premium consumer experiences, or multilingual deployments where the voice has to be indistinguishable from a human.
- Choose Bland if you want flat predictable per-minute pricing without component math, and your use case fits within the platform's prebuilt patterns.
- Choose AINORA if you do not want to assemble a voice AI from parts and would rather buy a working agent on a working phone number, fully managed.
Frequently Asked Questions
Frequently Asked Questions
Retell advertises $0.07 to $0.31 per minute. That total is the sum of three components: LLM cost ($0.003 to $0.080 per minute), voice infrastructure ($0.055 per minute flat), and text-to-speech ($0.015 to $0.040 per minute). A realistic mid-tier configuration with a good LLM and a premium voice lands around $0.10 to $0.18 per minute before extras like phone numbers, concurrency slots, and knowledge bases.
Beyond the per-minute rate, Retell charges $2 per phone number per month, $8 per concurrent call slot beyond the 20 included on pay-as-you-go, $8 per knowledge base beyond the first 10, and $10 per verified phone number per month. Branded calls and PII removal are additional add-ons. None of this is hidden in a deceptive sense - it is on the pricing page - but it is easy to miss when comparing the headline per-minute rate against competitors.
Both use a per-component per-minute model. Retell tends to land lower than Vapi in like-for-like configurations, mostly because the voice-infrastructure component is bundled into a single $0.055 line. Bland has publicly argued that competitor stacks land at $0.23 to $0.33 per minute against their own $0.11 to $0.14 flat. Run the math for your specific configuration before committing.
Yes, on the Enterprise tier with a Business Associate Agreement (BAA). HIPAA is not available on the pay-as-you-go plan. If you handle protected health information, you will need to engage Retell sales for an Enterprise quote with custom compliance terms.
20 concurrent calls are included on the pay-as-you-go plan. Additional concurrency costs $8 per slot per month. Enterprise plans include unlimited concurrency on dedicated infrastructure.
Retell ships native integrations with Make, Twilio, Vonage, Go High Level, n8n, and HubSpot. Anything beyond that is custom engineering work via Retell's API and webhooks.
Pick a developer platform like Retell, Vapi, ElevenLabs Conversational AI, or Bland if you have engineering capacity, want fine-grained control over LLM and TTS choices, and are building a custom voice product. Pick a managed voice AI service like AINORA if you want a working agent on a working phone number without assembling and operating the stack yourself. The trade-off is control versus time-to-value.
Founder & CEO, AInora
Building AI digital administrators that replace front-desk overhead for service businesses across Europe. Previously built voice AI systems for dental clinics, hotels, and restaurants.
View all articlesReady to try AI for your business?
Hear how AInora sounds handling a real business call. Try the live voice demo or book a consultation.
Related Articles
AInora vs Retell AI: Full-Service AI Receptionist vs Developer Toolkit
Managed AI receptionist vs developer API toolkit compared. Target user, time to value, integrations, and real costs.
Voicify AI Review 2026: Dental AI Receptionist for DSOs
Voicify AI dental receptionist review. Features, Denticon integration, DSO multi-location capabilities, pricing, pros and cons, and the top alternatives for 2026.
Can AI Really Talk Like a Human? Voice AI Explained
How modern voice AI achieves natural conversation, what is still hard, and what to listen for when evaluating vendors.