Quick answer: Vapi charges $0.05/min hosting plus speech-to-text, model and voice at provider prices, with your carrier billing call minutes separately; Retell AI charges $0.055/min plus its own LLM, voice and telephony rates. For 10,000 US outbound minutes (OpenAI model, ElevenLabs, Twilio) we get $959.85–$1,431.15 on Vapi and $1,244.50–$1,564.50 on Retell. Above 4 concurrent calls Vapi charges for capacity ($10 per extra line) where Retell includes 20, and HIPAA costs $2,000/month on Vapi.
Every price below was read from Vapi’s and Retell AI’s own pricing pages and documentation on 2 October 2026, in US dollars. Where a price depends on which speech-to-text (STT), large language model (LLM) or text-to-speech (TTS) provider you choose, the page says so and shows the range rather than inventing one number. The stack priced is Deepgram transcription (Vapi only), an OpenAI model, ElevenLabs voices and Twilio.
Vapi vs Retell AI pricing: what each one actually charges for
The two vendors build the same bill out of different parts.
Vapi is a hosting fee plus pass-through. Its pricing page describes the per-minute cost as “Four components: speech-to-text, the language model, text-to-speech, and transport. Each is billed at the provider’s listed price with no markup. Vapi’s $0.05/min hosting fee is added on top.” Transport (the phone network) is labelled “Charged by provider, not Vapi”, so if you call over Twilio, Twilio bills you for the minutes. On top of usage you can add an optional Success Package (Core, Pro or Premier) that changes support, service-level agreements and included capacity, and separate add-ons for extra concurrent call lines, extra workspaces and HIPAA.
Retell AI is a platform rate plus a menu of its own component prices. Its pricing page lists Retell Voice Infra at $0.055/minute, then a TTS price per voice family, an LLM price per named model, and a telephony price per country. It does not list a separate transcription line; Retell’s billing documentation describes Voice Infra as “the flat per-minute platform cost of running the real-time call, separate from LLM and telephony costs.” Retell quotes the whole pay-as-you-go range as “$0.07-$0.31 / min for AI Voice Agents”.
The practical consequence: on Vapi, the model line is a token bill that moves with your prompt; on Retell, the model line is a fixed per-minute price for each named model, with one documented exception for long prompts covered further down.
| Component | Vapi | Retell AI |
|---|---|---|
| Platform fee per minute | $0.05 hosting | $0.055 Voice Infra |
| Speech-to-text | Pass-through; calculator bands: Deepgram $0.00953–$0.00994, AssemblyAI $0.00521–$0.00747, OpenAI $0.0118–$0.01221 | No separate line on the pricing page |
| LLM | Pass-through, token-based; calculator bands: OpenAI $0.00774–$0.04523, Anthropic $0.00818–$0.03656, Google $0.006–$0.03329 | Per named model, e.g. GPT 5 mini $0.008, GPT 5.1 $0.04, GPT 4.1 $0.045, GPT 5.4 $0.080 (standard tier) |
| Text-to-speech | Pass-through; ElevenLabs $0.0146–$0.02383, Vapi voices $0.00695–$0.01934, Cartesia $0.00689–$0.01294 | $0.015 for Retell, Minimax, Fish, Cartesia, OpenAI and Inworld voices; ElevenLabs $0.040 |
| Telephony (US) | Charged by your carrier; page lists Twilio outbound $0.014, Twilio inbound $0.008, Vonage $0.00814, Telnyx $0.0055; Vapi telephony/SIP free | $0.015 US (Twilio); no charge for SIP trunking or custom telephony |
| Phone numbers | 1 free Vapi number on Usage only (5 on Core, 10 on Pro), inbound only and US only | $2/month (US, Canada); US toll-free $5/month |
| Concurrent calls included | 4 (Usage only), 10 (Core), 30 (Pro), custom (Premier) | 20 on pay-as-you-go |
| Extra concurrency | $10 per line per month | $8 per concurrent call per month |
| Packages | Core $29/month; Pro 10% of Vapi hosting fee, $999/month minimum; Premier custom | Enterprise: custom pricing |
| HIPAA | $2,000/month add-on, per organisation | BAA self-signed; “no additional fee to sign” |
| Outbound dialler fee | None listed on the pricing page | Batch Call +$0.005 per dial |
| Free credit | $5 | $10 |
Two cells in that table catch people out. First, Vapi’s “Vapi Telephony / SIP: Free” is true, but Vapi’s free-number documentation says “Free Vapi numbers are inbound only. Import a number from a supported telephony provider to make outbound calls.” For an outbound programme you are therefore paying a carrier on Vapi, always. Second, Retell’s cheap voice row is $0.015, but ElevenLabs voices on Retell are $0.040, and that single row is the biggest reason Retell’s bill below comes out higher on this particular stack.
What does 10,000 outbound minutes a month cost on Vapi vs Retell?
This is the worked calculation. Substitute your own minutes and the arithmetic still holds, because every line is a rate multiplied by a quantity you control.
Scenario A assumptions, stated in full:
- 10,000 connected, billed minutes a month, all US to US, outbound.
- Peak concurrency of 4 calls or fewer, so neither vendor charges for capacity. (10,000 minutes spread over 22 weekdays of 8 hours is about 0.95 calls in progress at any moment on average.)
- One local caller-ID number. On Vapi that number is a Twilio local number at Twilio’s own $1.15/month, because the free Vapi number cannot dial out.
- Vapi stack: Deepgram transcription, an OpenAI model, ElevenLabs voice, Twilio outbound at the $0.014/min Vapi lists (Twilio’s own page shows the same $0.0140/min for US local outbound). Vapi’s calculator publishes a band per provider, not a price per model, so the Vapi column carries a low and a high.
- Retell stack: Retell Voice Infra, ElevenLabs voice ($0.040), Retell US telephony ($0.015), one Retell number ($2/month), and two named OpenAI models from Retell’s table: GPT 5 mini ($0.008) and GPT 5.1 ($0.04).
- Outbound is run through each vendor’s own campaign tool. Retell charges $0.005 per dial for Batch Call; we assume an average connected call of 2 minutes (5,000 connected calls) and 2.5 dials per connected call, so 12,500 dials.
- No Success Package on Vapi, no Enterprise plan on Retell, no HIPAA, prompts under 4,000 tokens, no optional add-ons (knowledge base, denoising, branded caller ID).
| Line item | Vapi (low) | Vapi (high) | Retell, GPT 5 mini | Retell, GPT 5.1 |
|---|---|---|---|---|
| Platform: 10,000 × $0.05 (Vapi) / × $0.055 (Retell) | $500.00 | $500.00 | $550.00 | $550.00 |
| Speech-to-text: Deepgram band on Vapi | $95.30 | $99.40 | not a separate line | not a separate line |
| LLM: OpenAI band on Vapi; named model on Retell | $77.40 | $452.30 | $80.00 | $400.00 |
| Voice: ElevenLabs | $146.00 | $238.30 | $400.00 | $400.00 |
| Telephony per minute (billed by Twilio on Vapi) | $140.00 | $140.00 | $150.00 | $150.00 |
| Phone number, 1 per month | $1.15 | $1.15 | $2.00 | $2.00 |
| Dialler: 12,500 dials × $0.005 | $0.00 | $0.00 | $62.50 | $62.50 |
| Concurrency (4 or fewer) | $0.00 | $0.00 | $0.00 | $0.00 |
| Monthly total | $959.85 | $1,431.15 | $1,244.50 | $1,564.50 |
| Effective cost per minute | $0.0960 | $0.1431 | $0.1244 | $0.1564 |
How we recomputed it. Once line by line, as above, and once by summing each vendor’s per-minute rates first and adding the monthly items afterwards. For Vapi low that is ($0.05 + $0.00953 + $0.00774 + $0.0146 + $0.014) × 10,000 + $1.15 = $959.85; for Retell with GPT 5 mini it is ($0.055 + $0.008 + $0.040 + $0.015) × 10,000 + $2.00 + $62.50 = $1,244.50. Both methods agree to the cent on all four columns. As a third check, Vapi’s own calculator shows “$82 – $129 /mo” for 1,000 minutes on the same three providers without transport; our rates give $81.87–$129.00 for 1,000 minutes, which rounds to the vendor’s figure.
What Scenario A says. On this stack, Vapi’s band runs from $959.85 to $1,431.15 and Retell lands at $1,244.50 or $1,564.50 depending on the model. The ranges overlap, so the honest reading is not “Vapi is cheaper” but “Vapi’s low end is lower, and most of the gap is the voice line”: ElevenLabs costs $400.00 per 10,000 minutes on Retell against $146.00–$238.30 on Vapi. Swap Retell to one of its $0.015 voices and that line falls to $150.00, which puts Retell with GPT 5 mini at $994.50, inside Vapi’s band. The LLM row is not like-for-like either: Vapi’s figure is a provider band built on Vapi’s own token assumptions, Retell’s is a fixed price for a named model.
If you want to understand why the Vapi model line can drift upwards over a long call, our breakdown of voice AI cost per minute for realtime and pipeline stacks walks through how re-sent conversation history makes minute eight dearer than minute one.
Vapi vs Retell cost at 50,000 minutes and 30 concurrent calls
Scenario B is the same stack at a volume where capacity starts to cost money. The assumptions change in four places: 50,000 minutes a month, a peak of 30 simultaneous calls (a compressed calling window), 10 local caller-ID numbers, and 62,500 dials on Retell’s Batch Call at the same 2.5 dials per 2-minute connected call.
On Vapi, 30 lines can be bought three ways: Usage only plus 26 extra lines at $10 ($260), Core at $29 plus 20 extra lines ($229), or Pro, which includes 30 lines but costs 10% of the Vapi hosting fee with a $999 monthly minimum ($999, because 10% of $2,500 is $250). Core plus add-ons is the cheapest, so we use $229. On Retell, 20 calls are included and the other 10 cost $8 each, so $80.
| Line item | Vapi (low) | Vapi (high) | Retell, GPT 5 mini | Retell, GPT 5.1 |
|---|---|---|---|---|
| Platform | $2,500.00 | $2,500.00 | $2,750.00 | $2,750.00 |
| Speech-to-text (Deepgram band on Vapi) | $476.50 | $497.00 | not a separate line | not a separate line |
| LLM | $387.00 | $2,261.50 | $400.00 | $2,000.00 |
| Voice: ElevenLabs | $730.00 | $1,191.50 | $2,000.00 | $2,000.00 |
| Telephony per minute | $700.00 | $700.00 | $750.00 | $750.00 |
| Phone numbers, 10 per month | $11.50 | $11.50 | $20.00 | $20.00 |
| Dialler: 62,500 dials × $0.005 | $0.00 | $0.00 | $312.50 | $312.50 |
| Capacity for 30 concurrent calls | $229.00 (Core + 20 lines) | $229.00 (Core + 20 lines) | $80.00 (10 × $8) | $80.00 (10 × $8) |
| Monthly total | $5,034.00 | $7,390.50 | $6,312.50 | $7,912.50 |
| Effective cost per minute | $0.1007 | $0.1478 | $0.1262 | $0.1582 |
The same two-method recomputation agrees on all four totals. The capacity line now favours Retell by $149 a month, but the ElevenLabs line still costs Retell $808.50–$1,270.00 more than Vapi’s band at this volume, so on an ElevenLabs stack Vapi’s low end stays lower. Change the voice and the answer changes, which is why the threshold table below matters more than either total.
Core also brings a two-business-day response service-level agreement and 30 days of raw data retention, so the $229 is not pure capacity.
Where the Vapi and Retell bills diverge
Totals hide the switches. Find the row that matches your situation to see which vendor’s bill moves, and by how much.
| If your situation is… | Vapi | Retell AI | Which bill moves |
|---|---|---|---|
| Peak concurrency 1–4 | Included (Usage only) | Included (20 free) | Neither |
| Peak concurrency 5–6 | $10–$20 (extra lines) | $0 | Vapi +$10–$20 |
| Peak concurrency 7–20 | $29 (Core, 10 lines) to $129 (Core + 10 lines) | $0 | Vapi +$29–$129 |
| Peak concurrency 21–100 | $139–$929 (Core + lines) | $8–$640 | Vapi higher by $2 × lines + $89 (from $131 at 21 lines to $289 at 100) |
| You need HIPAA and a BAA | +$2,000 per organisation; HIPAA-compatible providers only | $0 to sign the BAA | Vapi +$2,000 |
| You want ElevenLabs voices | $0.0146–$0.02383/min | $0.040/min | Retell +$161.70 to +$254.00 per 10,000 min |
| You accept a budget voice (Cartesia) | $0.00689–$0.01294/min | $0.015/min | Retell +$20.60 to +$81.10 per 10,000 min |
| You bring your own SIP trunk or carrier | Unchanged: transport was always billed by the carrier | Retell’s $0.015/min and $2 numbers drop to $0; you pay your carrier | Retell −$150 per 10,000 min, before your carrier’s rate |
| You dial through the vendor’s campaign tool | No per-dial fee listed on vapi.ai/pricing as at 2 October 2026 | $0.005 per dial | Retell +$62.50 per 12,500 dials |
| Prompt above 4,000 tokens | Token bill rises with prompt and history | Call duration × (prompt tokens ÷ 4,000), rounded up | Both rise; Retell’s by a published formula |
| Traffic briefly exceeds your concurrency | Call cannot start until a slot frees | Optional burst: +$0.10/min for the whole call | Retell can cost more, but the call connects |
| You need a response SLA | Core $29 (2 business days); Pro $999 minimum (1 business day, 99% uptime) | Enterprise, custom pricing | Depends on quote |
| Calls go to Australia | Free numbers are US-only; import a number and pay that carrier | Lists Australia (Twilio) at $0.10/min; managed numbers are US and Canada only | Depends on your carrier rate |
Read the table top to bottom and two rows dominate. HIPAA is the largest single switch on either bill: $2,000 a month is more than the whole of Scenario A on Vapi’s low end. The voice row is the largest per-minute switch. Concurrency is real but small by comparison, never more than $289 a month apart at 100 lines. Retell says its Enterprise plans “offer custom concurrency starting at 50+ calls”, and its documentation says the dashboard shows “how high you can go”, so very high pay-as-you-go concurrency may need a conversation either way.
The bring-your-own-carrier row deserves a sentence of its own. On Retell, custom telephony removes Retell’s telephony line entirely (“Custom telephony (SIP trunking) carries no Retell telephony charge”), so the saving depends on what your carrier charges you per minute, which we have not priced here. On Vapi nothing changes, because Vapi never billed the carrier minutes in the first place. If you are choosing that carrier, our comparison of Twilio vs Telnyx costs for AI voice agents covers the carrier leg on its own.
If you would rather not assemble a voice stack at all: Zian AI runs autonomous AI sales agents across phone, SMS, email and WhatsApp, with SmartReach AI pacing the follow-ups and PrecisionPitch AI split-testing scripts, in 30+ languages. It is in partnership-application beta.
The billing rules that move the bill after launch
Both vendors publish rules that change the minutes you are billed for, and no calculator shows them.
Retell rounds to the second and skips failed calls. Its pricing FAQ says “Each call is tracked to the nearest second” with “No rounding up per call”, and “You are only charged for connected calls. If a call fails to connect, you won’t be billed. For calls that reach voicemail, billing applies only for the duration the AI agent is active on the line.” On transfers: “Once the call is transferred, the AI voice agent fee stops. Only the telephony fee continues for the remainder of the transferred call.”
Retell has two published exceptions. Its billing exceptions page sets a 10-second minimum on calls under 10 seconds that use a dynamic opening message when the AI speaks first, and scales billed duration for prompts above 4,000 tokens: “Scaling Factor = Prompt LLM Tokens ÷ 4,000” and “Billed Duration = Original Duration × Scaling Factor (rounded up)”. Retell’s own example is a 60-second call with 4,800 tokens billed as 72 seconds. The token count includes the transcript, tool history and retrieved knowledge base content, so it grows during a call. The rule is headed “LLM Price Scaling”, but the formula is written as billed duration, and the page does not spell out whether the scaled duration applies to the LLM line only or to every per-minute line. If it applied to every line, a 6,000-token agent would see its whole per-minute bill multiplied by 1.5; if only to the LLM line, the effect at GPT 5 mini rates would be small. Retell’s Flex Mode documentation frames the effect as model cost (“Flex mode can significantly increase your LLM costs”) but also says the rule “can multiply your costs several times over”, so it does not settle the question either. Ask Retell which before you model a long-prompt agent.
Vapi’s model line is an estimate until the call happens. Vapi’s cost documentation explains that transcription is billed per minute of audio (its estimate counts both the caller’s and the assistant’s channels), voice per character spoken, and the model per million input and output tokens. Its estimate assumes 5 model requests per turn, 150 output tokens per minute and a 50% prompt-cache hit rate, and its limitations table says that for long calls “Actual cost is often higher” because “Growing conversation history is sent again with each request”. The same two-row table cuts the other way for large prompts on models that support caching: “Actual cost is often lower” because “Actual cache use is usually higher than the estimate assumes”. Vapi also supports your own provider keys; whether that lowers the total is covered in our piece on bring-your-own API keys and voice agent cost.
Retell bills silence. “Billing covers the entire duration of the call because the speech-to-text engine remains active and listening throughout, even during silence.” Long hold times are paid minutes.
On Vapi, Twilio rounds each call up to the minute. Because the carrier bills the minutes on Vapi, Twilio’s own rule applies: its billing help article says “any partial minutes (under 60 seconds) are rounded up to the next full minute.” The tables above price the Twilio line on exact minutes, so treat Vapi’s telephony row as a floor: if call lengths are spread evenly, rounding adds about half a minute per call, roughly $35 a month on Scenario A’s 5,000 calls and $175 on Scenario B’s 25,000. Retell’s pricing FAQ, quoted above, says “No rounding up per call”.
The sentence on each pricing page that cuts against this comparison
Here is the sentence on each vendor’s pages that most weakens what we have said.
Against “Vapi’s low end is lower”: Vapi’s own documentation says the Cost value in its model metrics “estimates costs for model comparison; it is not an exact quote”, its pricing documentation presents the calculator as a way to “estimate your monthly cost”, and its limitations table says that for long calls “Actual cost is often higher”. Our Vapi low column inherits that caveat, and a long-prompt, long-call agent will not sit at the bottom of the band. Vapi’s docs also note that “Success Packages are currently available only to new Vapi customers. Customers on legacy pricing keep their existing billing structure and terms,” so an existing Vapi account may not be on the prices above.
Against “Retell is cheaper for HIPAA and high concurrency”: Retell’s billing-exceptions rule, “Billed Duration = Original Duration × Scaling Factor (rounded up)”, means a heavy prompt can raise Retell’s effective rate, and its own documentation names Flex mode as “a common trigger”. A HIPAA saving of $2,000 a month can be eroded by an agent design that runs well above 4,000 tokens.
Against “Retell’s telephony is simple”: Retell’s country table lists Australia (Twilio) at $0.10/min, and Retell-managed numbers are US and Canada only, so the clean $0.015 US figure does not travel.
Who each pricing model is wrong for
Vapi’s pricing is the wrong fit if you handle protected health information and are small: the $2,000 monthly HIPAA add-on will outweigh every other line until volume is large. It also costs more for capacity whenever you need more than 4 simultaneous calls (from 21 lines up, $2 per line more than Retell plus an $89 base), and it never gives you one invoice for outbound, because the carrier bills you separately.
Retell’s pricing is the wrong fit if you are set on ElevenLabs voices at volume, where the $0.040 rate costs $161.70–$254.00 more per 10,000 minutes than Vapi’s band; if your agent design pushes prompts well above 4,000 tokens; or if your calls originate outside the US and Canada and you do not want to run custom telephony.
Price stops deciding it if you stay at 4 concurrent calls or fewer and use a budget voice. Rerunning Scenario A with Cartesia on Vapi and a $0.015 voice on Retell gives $882.75–$1,322.25 on Vapi against $994.50 (GPT 5 mini) or $1,314.50 (GPT 5.1) on Retell. Those ranges overlap, so developer experience, support terms and where your calls go matter more than the rate card.
For the wider product trade-offs rather than the bill, our Retell AI outbound comparison covers packaging and compliance posture in more depth; note that it is written by a vendor and says so.
What this page does not price
- Your carrier’s rate when you bring your own SIP trunk on either platform.
- Vapi’s per-model LLM prices: its calculator shows provider bands, not a price for each model.
- Enterprise or Premier contracts, which both vendors quote privately.
- Optional add-ons: Retell’s knowledge base (+$0.005/min), denoising (+$0.005/min), safety guardrails (+$0.005/min), PII removal (+$0.01/min), branded call (+$0.10 per outbound call) and verified numbers ($10/month); Vapi’s additional workspaces ($20/month) and data-retention options.
- Twilio charges beyond outbound minutes and number rental, such as recording or calls-per-second upgrades.
- Australian GST and currency conversion; every figure here is US dollars before tax.
- The compliance obligations of calling itself, which no platform price covers.
One method note, because this page sits on a vendor’s site: Zian AI does not publish a per-minute price during its partnership-application beta, so it is deliberately absent from every table above.
Where every figure on this page comes from
| Figure | Who published it | Link | Date read |
|---|---|---|---|
| $0.05/min hosting; components billed “with no markup” | Vapi | vapi.ai/pricing | 2 October 2026 |
| Calculator bands: Deepgram, AssemblyAI, OpenAI STT; OpenAI, Anthropic, Google LLM; ElevenLabs, Vapi, Cartesia voice | Vapi | vapi.ai/pricing | 2 October 2026 |
| “$82 – $129 /mo” estimate at 1,000 minutes | Vapi | vapi.ai/pricing | 2 October 2026 |
| Transport: Twilio outbound $0.014, inbound $0.008, Vonage $0.00814, Telnyx $0.0055; charged by provider | Vapi | vapi.ai/pricing | 2 October 2026 |
| Core $29/month; Pro 10% of hosting, $999 minimum; included lines 4/10/30; numbers 1/5/10; $5 credit | Vapi | vapi.ai/pricing | 2 October 2026 |
| Extra concurrency $10/line/month; workspaces $20/month; HIPAA $2,000/month per organisation | Vapi | vapi.ai/pricing | 2 October 2026 |
| Core 2-business-day SLA, 30-day retention; Success Packages for new customers only | Vapi | docs.vapi.ai/billing/pricing-and-success-packages | 2 October 2026 |
| Free Vapi numbers inbound only, US only | Vapi | docs.vapi.ai/free-telephony | 2 October 2026 |
| 5 requests per turn, 150 output tokens/min, 50% cache assumption; long calls often higher | Vapi | docs.vapi.ai/…/understanding-cost | 2 October 2026 |
| HIPAA mode limits providers; HIPAA and ZDR mutually exclusive | Vapi | docs.vapi.ai/security-and-privacy/hipaa | 2 October 2026 |
| US local outbound $0.0140/min; local number $1.15/month | Twilio | twilio.com/en-us/voice/pricing/us | 2 October 2026 |
| Partial minutes rounded up to the next full minute | Twilio | support.twilio.com article 223132307 | 2 October 2026 |
| Flex mode “can significantly increase your LLM costs”; token rule “can multiply your costs several times over” | Retell AI | docs.retellai.com/build/conversation-flow/flex-mode | 2 October 2026 |
| Voice Infra $0.055/min; TTS $0.015 and ElevenLabs $0.040; $0.07–$0.31/min range | Retell AI | retellai.com/pricing | 2 October 2026 |
| LLM: GPT 5 mini $0.008, GPT 5.1 $0.04, GPT 4.1 $0.045, GPT 5.4 $0.080 per minute | Retell AI | retellai.com/pricing | 2 October 2026 |
| Telephony US (Twilio) $0.015/min, Australia (Twilio) $0.10/min; no charge for custom telephony | Retell AI | retellai.com/pricing | 2 October 2026 |
| Numbers $2/month; 20 free concurrent calls, then $8/month; Enterprise from 50+ calls; $10 free credit | Retell AI | retellai.com/pricing | 2 October 2026 |
| Batch Call $0.005/dial; knowledge base, denoising, guardrails $0.005/min; PII removal $0.01/min; branded call $0.10; verified number $10/month | Retell AI | retellai.com/pricing | 2 October 2026 |
| Per-second billing; connected calls only; transfer and silence rules | Retell AI | retellai.com/pricing | 2 October 2026 |
| US toll-free number $5/month; managed numbers US and Canada only | Retell AI | docs.retellai.com/deploy/purchase-number | 2 October 2026 |
| 10-second minimum; 4,000-token scaling formula and 4,800-token example | Retell AI | docs.retellai.com/accounts/billing-exceptions | 2 October 2026 |
| Burst surcharge $0.10/min for the entire call | Retell AI | docs.retellai.com/deploy/concurrency | 2 October 2026 |
| BAA self-signed, no additional fee | Retell AI | docs.retellai.com/general/compliance | 2 October 2026 |
| Voice Infra definition; custom telephony carries no Retell telephony charge | Retell AI | docs.retellai.com/accounts/billing | 2 October 2026 |
| Scenario totals, per-minute figures, concurrency differences, 12,500 and 62,500 dials | Zian AI calculation from the rows above | This page | 2 October 2026 |
Frequently asked questions
Is Vapi cheaper than Retell AI?
It depends on the voice and the concurrency. For 10,000 US outbound minutes on an OpenAI model, ElevenLabs and Twilio, we calculate $959.85–$1,431.15 a month on Vapi and $1,244.50–$1,564.50 on Retell AI, with most of the gap from ElevenLabs costing $0.040/min on Retell. Above 4 concurrent calls Retell’s capacity line is the cheaper one, and HIPAA adds $2,000 a month on Vapi with no fee to sign Retell’s BAA.
What is Vapi’s cost per minute compared with Retell’s?
Vapi charges a $0.05/min hosting fee and passes STT, model and voice costs through at provider prices, while your carrier bills the call minutes directly. Retell AI charges $0.055/min for Voice Infra plus per-model LLM, voice and telephony rates, and quotes $0.07–$0.31/min overall. Our Scenario A stacks (10,000 minutes) land at $0.0960–$0.1431/min on Vapi and $0.1244–$0.1564/min on Retell; at 50,000 minutes and 30 concurrent calls (Scenario B) they land at $0.1007–$0.1478/min and $0.1262–$0.1582/min.
How much do extra concurrent calls cost on Vapi and Retell?
Vapi includes 4 concurrent calls on Usage only, 10 on Core and 30 on Pro, and sells extra lines at $10 per line per month. Retell AI includes 20 on pay-as-you-go and charges $8 per extra concurrent call per month. At 30 lines that is $229 on Vapi (Core plus 20 lines) against $80 on Retell.
How much does HIPAA cost on Vapi vs Retell?
Vapi lists HIPAA-eligible data handling as a $2,000/month add-on per organisation, with a standard DPA and BAA. Retell AI’s security and compliance documentation says its BAA is available for self-signing and that there is no additional fee to sign it. On Vapi, HIPAA mode also limits you to compatible providers.
Can I use my own Twilio number or SIP trunk with Vapi or Retell?
Yes, on both. Vapi’s free numbers are inbound only, so outbound on Vapi always uses an imported number, and the carrier bills the minutes. On Retell AI, custom telephony removes Retell’s $0.015/min US telephony line and $2 number fee; you pay your carrier instead.
Do Vapi or Retell give free credit to start?
Vapi’s pricing page lists $5 in free credits on Usage only. Retell AI lists $10 in free credits, and its billing documentation says the trial credit is granted once per email address.
Does Retell charge for unanswered calls or voicemail?
Retell AI says you are only charged for connected calls, and for calls that reach voicemail, billing applies only while the AI agent is active on the line. Separately, a call under 10 seconds that uses a dynamic opening message when the AI speaks first is billed as 10 seconds.
Running outbound and weighing build vs buy? Zian AI’s autonomous sales agents handle calls, SMS, email and WhatsApp follow-up with CRM integrations for HubSpot, Salesforce and HighLevel. Zian is taking partners during its partnership-application beta.