Retell AI, Vapi and ElevenLabs side by side
| Compared on | 01Retell AI | 02Vapi | 03ElevenLabs |
|---|---|---|---|
| Typical cost | $0.07 to $0.31 a minute pay-as-you-go, depending on voice, model, telephony and add-ons | $0.05 a minute for Vapi hosting, plus speech, model and carrier costs | Plans from $6 a month with minutes included; extra minutes $0.08; LLM usage billed on top |
| Phone lines | Twilio or Telnyx at $0.015 a minute, or your own SIP trunk at no charge | Vapi telephony and SIP free; carrier charges billed by the provider | Telephony included on every plan; native Twilio and a SIP trunk |
| Security and compliance | Says it is SOC 2 Type 1 and Type 2 certified and signs a BAA for healthcare | HIPAA add-on at $2,000 a month, with a BAA; zero data retention from Core | BAAs for HIPAA customers on Enterprise |
| Response time (vendor's claim) | About 600 milliseconds | Under 600 milliseconds | No figure published on its pricing or overview pages |
| Main risk | Per-minute add-ons that push up the real rate | More choices to make and test before launch | Plan minutes that do not match your real call volume |
All figures come from each vendor's own pages, read on 17 November 2026. Per-minute totals depend on the voice, model and phone line you pick, so price a pilot with your own call mix.
When to choose Retell AI
Choose Retell AI when you want a phone agent for bookings, reception or follow-up calls and you would rather configure than code. Retell lists appointment booking, warm transfers to a person, and integrations with HubSpot, Go High Level, Make and n8n.
Pricing is itemised. Voice infrastructure is $0.055 a minute, then you add a voice, a model and telephony. Add-ons such as a knowledge base or PII removal are charged per minute on top.
Concurrency starts generous. Every account gets 20 concurrent calls free; more cost $8 per line a month, and an optional burst lets overflow calls connect at an extra $0.10 a minute.
Compliance is stated up front. Retell says it is SOC 2 Type 1 and Type 2 certified, HIPAA and GDPR compliant, and will sign a BAA for healthcare workloads.
When to choose Vapi
Choose Vapi when you have developers and want control over each part of the call. Vapi describes itself as a developer platform for building voice agents, with a choice of speech, model and voice providers.
Hosting and models are billed apart. Vapi charges $0.05 a minute for hosting and says it passes model costs through with no markup.
Several assistants can share a call. Squads hand a caller between specialised assistants while keeping the context.
Plans buy capacity. The free start includes 4 concurrent calls and 1 number; Core ($29 a month) has 10 and 5; Pro (10% of the hosting fee, $999 a month minimum) has 30 and 10. Extra lines are $10 each a month.
Regulated data costs extra. HIPAA-eligible handling with a BAA is a $2,000-a-month add-on.
The flexibility is the point, and it also means your team owns more of the testing.
When to choose ElevenLabs
Choose ElevenLabs when the voice itself is the deciding factor: a brand voice, a natural accent or several languages. Its agent product, ElevenAgents, includes telephony on every plan, a native Twilio integration, a SIP trunk and batch outbound calls.
Minutes come bundled. Starter ($6 a month) includes 75 minutes, Pro ($99) 1,238 and Business ($990) 12,375. Extra minutes cost $0.08, and LLM usage is billed separately on top.
Busy periods are covered. Burst lets agents take up to three times their concurrency limit, with the excess billed at double the standard rate.
Knowledge is built in. You can upload documents and turn on RAG so the agent answers from your material.
Healthcare needs Enterprise. BAAs for HIPAA customers are listed on the Enterprise plan.
When a voice agent is the wrong tool
A voice agent is the wrong tool when the call needs judgment, empathy or authority that a script cannot give.
Hard conversations. Complaints, bereavement, debt and medical worries need a person on the line, or at least a fast route to one.
Low volume. If you get a handful of calls a day, a call-back form, a shared inbox or an answering service may cost less and annoy fewer people.
A back office that is not ready. If bookings live in a paper diary or a calendar nobody updates, the agent will promise slots that do not exist. Fix the calendar first.
Rules you have not checked. Laws on call recording, consent and telling callers they are speaking to AI differ by country and state. This guide is not legal advice; check with a lawyer before launch.
What goes wrong six months after go-live
A phone agent rarely breaks loudly; it drifts. These are the things to watch.
The real cost per minute creeps up. Long prompts, add-ons, transfers and silence all count. Retell, for example, bills silence and hold time and increases billed time when prompts average over 4,000 tokens. Compare invoices with call logs every month.
Concurrency limits bite on busy days. Know what happens to the next caller when every line is busy, and whether burst is switched on.
Integrations go stale. A calendar, a CRM field or a price changes and the agent books or quotes the wrong thing.
Nobody listens to the calls. Sample recordings every week and keep a list of questions the agent got wrong.
The model underneath changes. Providers retire and replace models. Re-test the agent before you switch, not after.
How we approach it at Vizio AI
We start by listening to your current calls, or reading the call notes, and sorting them into three groups: what an agent can handle fully, what it should pass to a person, and what it should never touch. Then our DX Studio picks the platform that fits your team, connects it to your calendar and CRM, writes the transfer rules, and tests it with real scenarios before any customer hears it. After launch we review recordings with you and adjust the agent.
FAQs
Sources
- Retell AI: Pricingper-minute range and components, concurrency, telephony and billing rules; read 17 November 2026
- Retell AI: Homepageresponse time claim, booking, transfers, integrations, compliance claims
- Vapi: Pricinghosting fee, plans, concurrency, phone numbers, HIPAA add-on; read 17 November 2026
- Vapi Docs: Introductiondeveloper platform, squads, response time claim
- ElevenLabs: ElevenAgents pricingplans, included minutes, extra and burst rates, telephony, LLM billing; read 17 November 2026
- ElevenLabs Docs: ElevenAgents overviewTwilio, SIP trunk, batch calls, tools and RAG
