Retell AI vs Vapi, compared.
Retell AI is a dashboard for configuring phone agents, with API access for custom integration. Vapi is an API and SDK for building one in code. Neither one starts from a plain-language brief.
Where each fits.
Retell AI
Retell AI is a voice AI platform built around a dashboard for configuring phone agents, with API access for custom integrations and an on-premises option for enterprise teams.
Vapi
Vapi is a developer-facing voice AI platform, built API-first: an API and SDKs for wiring large language models, speech providers, and call logic into code, then deploying through Vapi's own infrastructure.
How the two compare.
Six differences from what both sides publish.
| Aspect | Retell AI | Vapi |
|---|---|---|
| How you build it | Built through Retell AI's dashboard, with API access for custom integration. | Configured through Vapi's API and SDKs, in code. |
| Where it runs | Retell AI's cloud dashboard, with an on-premises option for enterprise. | Vapi's own hosting, free to start; a $29/month Core plan and a $999/month-minimum Pro plan for higher volume. |
| How the price is put together | A pay-as-you-go per-minute range built from separate components: a Retell Voice Infra rate, text-to-speech, a per-model language-model rate, and telephony. | A platform-hosting fee plus separate, per-minute rates for the model, transcription, voice, and telephony, each from its own provider. |
| Concurrent calls | 20 concurrent calls included on the pay-as-you-go plan, then $8/month for each additional concurrent call; Enterprise states no cap. | 4 concurrent calls free; 10 on the $29/month Core plan; 30 on the $999/month-minimum Pro plan; extra concurrency is $10 per line a month. |
| Testing before live | A live demo and free trial, per Retell AI's own site. | Evals, deterministic pass/fail checks on a single decision point, and Simulations, a full conversation run by an AI tester that adapts and is graded on outcome. Vapi's own docs present the two as complementary, not one replacing the other. |
| Who it's for | Teams who want a configurable dashboard for call flows. | Developers building a voice agent in code, on their own stack. |
What each costs, by layer.
Read straight off retellai.com/pricing and vapi.ai/pricing on 24 September 2026. Where a page doesn't publish a rate for a layer, the cell is the em dash, never an estimate.
| Layer | Unit | Retell AI | Vapi |
|---|---|---|---|
| Carrier (telephony) | per minute | $0.015/min through Retell, varying by country and provider; free with your own SIP trunk. | Free on Vapi’s own SIP trunk; $0.008/min inbound and $0.014/min outbound via Twilio, $0.0055/min via Telnyx, $0.00814/min via Vonage. |
| Transcription (STT) | per minute | — | $0.0095 to $0.0099/min (Deepgram). |
| Model (LLM) | per minute | $0.0016 to $0.32/min, priced per model across Retell’s published list. Mid-tier examples: GPT-5.6 Terra and Claude 5 Sonnet at $0.064/min, GPT-5.5 at $0.16/min. | $0.0077 to $0.0452/min (OpenAI). |
| Voice (TTS) | per minute | $0.015/min on Retell’s own, Minimax, Fish, Cartesia, OpenAI, or Inworld voices; $0.040/min on ElevenLabs. | $0.0146 to $0.0238/min (ElevenLabs). |
| Platform (per-minute base rate) | per minute | $0.055/min, billed as “Retell Voice Infra” on every call, on top of the model, voice, and telephony rates above. | $0.05/min pay-as-you-go and on the Core plan; on the Pro plan, 10% of the Vapi hosting fee, $999/month minimum. |
| Monthly plan fee | per month | $0 on pay-as-you-go, with $10 in free credit; Enterprise is custom. | $0 to start; $29/month on Core; $999/month minimum on Pro; Premier is custom. |
| Phone number rental | per month | $2/month per phone number; $10/month for a verified number. | — |
Sources: retellai.com — homepage (checked 2026-08-30) · retellai.com/pricing (checked 2026-09-24) · vapi.ai — homepage (checked 2026-08-30) · vapi.ai/pricing (checked 2026-09-24) · docs.vapi.ai/test/voice-testing (checked 2026-09-24)
Where a Callable fits.
Retell AI and Vapi are both platforms you build on: Retell AI through a dashboard for configuring phone agents, with API access for custom integration; Vapi through an API and SDK integration, in code. A Callable skips building altogether. Describe the line in plain language, the same way you'd brief a person, and the builder asks what it can't infer, then drafts and tests it before anything goes live.
Every call has an itemized receipt: carrier, transcription, model, and voice, each with its own source and rate, rather than a per-minute range built from separate components or a hosting fee plus separately billed provider lines.
Where to go next.
Describe your line
Tell the builder what calls it should take. It asks what it can't infer, then drafts the line.
Start building →See what it costs
Every call prints a receipt with the source and rate for every layer.
View cost →Watch how it’s built
From brief to published line.
See how it works →
Also compare: Retell AI · Vapi · Bland AI vs Vapi · All platforms compared
Straight answers about Retell AI and Vapi.
Six questions people ask when choosing between Retell AI and Vapi.
No. Retell AI vs Vapi is a real choice if you want to build a phone agent yourself, through a dashboard or in code. A Callable is a third path: describe the line in plain language, and the builder drafts, tests, and publishes it, with no dashboard flow to rebuild and no API to wire up.
It depends on volume and what you count. Retell AI publishes a pay-as-you-go range of $0.07 to $0.31 a minute for voice agents, built from a $0.055/min Voice Infra rate, text-to-speech at $0.015 to $0.040/min, a language model priced per model from $0.0016 to $0.32/min across its published list, and telephony from $0.015/min. Vapi charges $0.05/min for its own hosting (or a $29 to $999-plus monthly plan) on top of separate model, transcription, voice, and telephony rates from their own providers. Add up both sides at your own call volume before comparing a single headline number.
Vapi does, an API and SDK integration. Retell AI doesn't require it to start: its dashboard configures phone agents directly, with API access available for custom integration. A Callable's builder asks what it can't infer and drafts the line directly, with no dashboard flow to maintain and no API to wire up.
A pay-as-you-go range of $0.07 to $0.31 a minute for voice agents: a $0.055/min Voice Infra rate, text-to-speech at $0.015/min on Retell’s own and partner voices or $0.040/min on ElevenLabs, and a language model priced per model, from $0.0016/min to $0.32/min across Retell’s published list. Telephony runs from $0.015/min, or free with your own SIP trunk. Twenty concurrent calls are included, then $8/month for each additional concurrent call; phone numbers are $2/month, or $10/month verified.
Vapi's own hosting fee, $0.05/min pay-as-you-go or folded into a $29/month Core or $999-minimum Pro plan, plus separate per-minute rates for the transcriber ($0.0095 to $0.0099), the model ($0.0077 to $0.0452), the voice ($0.0146 to $0.0238), and telephony (free on Vapi's own SIP trunk, or $0.0055 to $0.014/min through Twilio, Telnyx, or Vonage).
About $0.062 a minute on the default stack, at the providers’ published rates, itemized by layer on every call: carrier, transcription, model, voice. See /cost. Retell AI and Vapi price differently again: a per-minute range built from separate components on one side, a hosting fee plus separate provider rates on the other.
About $0.062 a minute on the default stack, at the providers’ published rates, itemized by layer on every call: carrier, transcription, model, voice. See the price book →
For the full pricing comparison across five platforms, Callable’s own price book, a calculator for your own call volume, and every alternative compared: The full pricing comparison across five platforms · The full price book · Run your own call volume · Every alternative, compared
Describe the line you need.