Vapi Pricing per Minute: The Full Voice AI Cost (2026)
Vapi pricing starts at $0.05 per connected call minute without a Success Package, but that is only Vapi's platform and hosting fee. The actual Vapi cost per minute also includes speech-to-text, LLM tokens, text-to-speech, and telephony. Under the named September 2026 assumptions in this guide, reproducible examples range from about $0.071 per minute for a low-cost web configuration to about $0.215 for a premium US outbound configuration with a large model context.
Vapi does not publish one universal all-in rate because customers choose the providers. That flexibility is one of its strengths, but it makes a configuration table more useful than a single headline number.
The Bottom Line
Vapi's real per-minute price is the sum of five independently billed layers.
- Vapi platform: $0.05 per connected minute without a Success Package.
- STT: billed by audio time, with Vapi estimating both caller and assistant audio channels.
- LLM: billed by input and output tokens, which grow with prompts, tools, and conversation history.
- TTS: billed by generated characters or the chosen voice provider's unit.
- Telephony: $0 for a modeled web call, a separate carrier bill for BYO telephony, or the applicable phone-network rate.
The $0.05 claim is accurate as a platform price. It is incomplete as a call price.
What Does Vapi Charge per Minute?
As of September 2026, Vapi charges $0.05 per minute for hosting without a Success Package and passes STT, LLM, and TTS through at provider cost. Customers can also bring their own model-provider API keys, in which case those providers bill them directly. Vapi also lists Core at $29/month and Pro at 10% of Vapi hosting, subject to a $999/month minimum, while Premier is custom.
The official Vapi pricing page separates the invoice this way:
| Public Vapi item | Published price |
|---|---|
| Platform and hosting for calls | $0.05/min |
| STT, LLM, and TTS | At provider cost, or billed through BYO keys |
| Included concurrency without a Success Package | 4 active call lines |
| Core plan | $29/month; 2 organisations, 5 numbers, 10 concurrent calls, 30-day retention, and ZDR |
| Pro plan | 10% of Vapi hosting, subject to a $999/month minimum; 10 organisations, 10 numbers, 30 concurrent calls, 180-day retention, RBAC, and ZDR |
| Premier plan | Custom |
| Additional concurrency | $10/line/month |
| HIPAA add-on | $2,000/month |
| Zero Data Retention | Included with Core and higher Success Packages |
| Call history retention | 14 days without a Success Package |
Larger plans and committed-volume agreements can change the economics. This article models the public $0.05 hosting rate, not a negotiated contract.
For a market-wide explanation of the same cost layers, use the voice AI pricing per minute guide.
How Is the Actual Vapi Cost per Minute Calculated?
The actual cost is platform plus transcription plus model tokens plus generated speech plus telephony. Optional features and fixed monthly add-ons sit above that call-level total.
Use this formula:
Vapi call cost/min = $0.05 + STT + LLM input + LLM output + TTS + telephony + per-minute add-ons
Then add monthly items:
Monthly total = call minutes x call cost/min + plan or success package + extra concurrency + phone numbers + compliance add-ons + other fixed services
Vapi's component-cost methodology explains why estimates vary. Transcription uses audio minutes, TTS uses spoken characters, and the LLM uses input and output tokens. Vapi estimates 150 output tokens per minute and calculates input from the assistant's actual prompt and tool definitions. Long calls can cost more because growing conversation history is repeatedly sent to the model.
What Assumptions Do the Cost Scenarios Use?
The three scenarios below use public first-party rates and expose every assumption. They are reproducible examples, not promises of what every Vapi assistant will cost.
Common assumptions:
- Vapi public hosting fee without a Success Package: $0.05 per minute.
- The call is connected for the full billed minute.
- STT covers two audio channels, following Vapi's methodology.
- The assistant produces 750 TTS characters and 150 LLM output tokens per minute.
- No knowledge-base, recording, compliance, concurrency, or other optional charges are included in the per-minute row.
- Taxes, contracted discounts, provider minimums, and number rental are excluded.
The source rates checked for this model are:
- Deepgram pricing: Nova-3 monolingual streaming at $0.0048 per audio minute and multilingual at $0.0058.
- OpenAI GPT-5 pricing: nano at $0.05 input and $0.40 output per million tokens, mini at $0.25 and $2, standard at $1.25 and $10.
- Deepgram TTS pricing: Aura-1 at $0.015 and Aura-2 at $0.030 per 1,000 characters.
- ElevenLabs API pricing: Multilingual v2/v3 at $0.10 per 1,000 characters.
- Twilio US voice pricing: local inbound at $0.0085 per minute and local outbound at $0.014 per minute.
Provider prices move independently. Recheck them before using the table in a client quote.
What Is a Low-Cost Vapi Web Configuration?
The low-cost web scenario totals about $0.0714 per minute. It uses economical speech and model components and no public telephone-network leg.
Named configuration:
- Vapi: $0.05000
- Deepgram Nova-3 monolingual STT:
2 x $0.0048 = $0.00960 - GPT-5 nano with 10,000 input and 150 output tokens:
$0.00050 + $0.00006 = $0.00056 - Deepgram Aura-1 with 750 characters:
750 / 1,000 x $0.015 = $0.01125 - Telephony: $0.00000 for the modeled browser or web call
Total: $0.07141 per minute
This scenario shows why Vapi appeals to technical teams. A compact prompt, economical model, inexpensive voice, and no PSTN carrier can produce a low raw rate while preserving control over every component.
What Is a Mid-Range Vapi US Inbound Configuration?
The mid-range US inbound scenario totals about $0.0934 per minute. It uses a stronger small model, Deepgram's current Aura-2 voice, and Twilio local inbound telephony.
Named configuration:
- Vapi: $0.05000
- Deepgram Nova-3 monolingual STT: $0.00960 for two channels
- GPT-5 mini with 10,000 input and 150 output tokens:
$0.00250 + $0.00030 = $0.00280 - Deepgram Aura-2 with 750 characters:
750 / 1,000 x $0.030 = $0.02250 - Twilio US local inbound: $0.00850
Total: $0.09340 per minute
This is still not a universal “typical Vapi price.” A longer system prompt, more tool definitions, provider add-ons, toll-free number, recording, or different voice changes the total.
What Is a Premium Vapi US Outbound Configuration?
The premium US outbound scenario totals about $0.2146 per minute. The largest change is a 50,000-token input context per minute combined with a premium multilingual voice.
Named configuration:
- Vapi: $0.05000
- Deepgram Nova-3 multilingual STT:
2 x $0.0058 = $0.01160 - GPT-5 with 50,000 input and 150 output tokens:
$0.06250 + $0.00150 = $0.06400 - ElevenLabs Multilingual v2/v3 with 750 characters:
750 / 1,000 x $0.10 = $0.07500 - Twilio US local outbound: $0.01400
Total: $0.21460 per minute
The 50,000 input-token assumption is intentionally large. It represents an assistant with substantial prompt, tools, or repeated conversation context, not an unavoidable Vapi cost. Vapi's own methodology identifies input tokens per minute as the largest cost factor customers can control.
What Does Vapi Cost at 1,000, 10,000, and 100,000 Minutes?
At the public $0.05 hosting rate, the three named scenarios range from $71.41 to $214.60 at 1,000 minutes and from $7,141 to $21,460 at 100,000 minutes. These totals exclude fixed add-ons, number rental, taxes, and volume contracts.
| Monthly minutes | Low web, $0.07141/min | Mid US inbound, $0.09340/min | Premium US outbound, $0.21460/min |
|---|---|---|---|
| 1,000 | $71.41 | $93.40 | $214.60 |
| 10,000 | $714.10 | $934.00 | $2,146.00 |
| 100,000 | $7,141.00 | $9,340.00 | $21,460.00 |
The Vapi platform portion alone is $50, $500, or $5,000 at those volumes. The rest belongs to the named speech, model, voice, and carrier configuration.
At 100,000 minutes, a committed-volume agreement may change the answer materially. Request current pricing rather than multiplying the public hosting fee and assuming no discount.
Are Vapi Web Calls Free?
Vapi web calls avoid the public telephone-network charge, but they are not free calls. The $0.05 Vapi platform fee and the selected STT, LLM, and TTS costs still apply.
The low scenario models telephony as $0 because the call remains in a browser or app rather than entering the PSTN. If the implementation uses another real-time media, WebRTC, or transport vendor that charges separately, add that vendor to the formula.
Web calls are often the cheapest way to test a Vapi assistant because they remove phone numbers and carrier minutes. They do not remove AI processing.
How Does Bring Your Own Telephony Affect Vapi Pricing?
Bring your own telephony moves the carrier charge to the agency's chosen provider or existing SIP arrangement. Vapi still bills the platform minute, and the speech and model providers still bill their components.
BYO telephony can be cheaper at scale or useful when the customer already has negotiated carrier rates. It can also add another invoice, routing configuration, and support boundary. Compare the carrier's origination, termination, phone-number, transfer, recording, and emergency-call fees, not just its lowest route price.
The Twilio rates in this article are examples for US local calling. Toll-free, international, browser, SIP, forwarding, and high-cost destinations have different prices.
What Are Vapi's Hidden Costs and Add-Ons?
Vapi's extra costs are mostly visible once the buyer stops treating $0.05 as the whole invoice. Model providers, telephony, concurrency, compliance, retention, and the agency product layer are separate decisions.
Concurrency
The no-Success-Package option lists 4 concurrent calls, while Core lists 10. Additional capacity is priced separately. Check the current plan table before projecting a high-concurrency campaign.
Concurrency is capacity, not usage. A high-volume appointment-reminder campaign may need more simultaneous lines than an inbound receptionist with the same monthly minutes.
HIPAA
Vapi lists HIPAA at $2,000 per month. Its HIPAA documentation also requires a signed BAA and compatible STT, LLM, and TTS providers. HIPAA mode and Zero Data Retention are mutually exclusive in Vapi; one must be disabled before the other is enabled.
Buying the add-on is not enough if a selected component cannot process protected health information under the required terms. Every provider in the pipeline must fit the compliance configuration.
Zero Data Retention
Vapi's current public pricing includes Zero Data Retention with Core and higher Success Packages. The option without a Success Package lists 14 days of call-history retention. ZDR cannot be enabled at the same time as HIPAA mode. Confirm the exact storage and provider configuration required for your workload, because downstream model and speech providers remain part of the data path.
Phone numbers and carrier features
Phone-number rental, toll-free rates, recording, answering-machine detection, forwarding, and other carrier features can add costs outside Vapi. For example, Twilio currently lists a US local number at $1.15 per month, recording at $0.0025 per minute, and answering-machine detection at $0.0075 per call.
White-label agency layer
Vapi does not publicly list a turnkey white-label agency portal with branded client workspaces and rebilling. Its APIs and organization features can support a custom agency product, or an agency can subscribe to a third-party wrapper. Add the development or wrapper cost, client-seat charges, and operating overhead to the Vapi stack.
The Trillet vs Retell vs Vapi comparison covers that architecture decision separately from the per-minute calculation.
When Is Vapi the Cheapest Choice?
Vapi can be the cheapest choice when a technical team uses compact prompts, economical models, low-cost voices, web or negotiated telephony, and already owns the surrounding product layer. The low web scenario demonstrates that advantage at about $0.071 per minute.
Vapi is also attractive when provider choice is the product. Teams can swap speech, model, and voice components, use their own keys, and optimize a specific workflow rather than accept one bundled stack.
The honest trade-off is operational. An agency that needs client branding, workspaces, rebilling, support processes, and compliance documentation must account for the cost of building or renting those capabilities.
When Is a Bundled White-Label Platform Cheaper?
A bundled platform can be cheaper when the buyer values predictable model costs and a finished agency layer more than component optimization. The raw per-minute rate may be higher than an aggressively optimized Vapi stack while the complete operating cost is lower.
Trillet White-Label charges $0.12 per AI minute for platform, STT, LLM, and TTS after included minutes. Telephony is separate: web and bring your own telephony add no Trillet telephony fee, US Trillet telephony starts at $0.014 per minute, international rates vary, and transferred time costs $0.05 per minute after AI usage stops.
That is not a claim that Trillet beats every Vapi configuration. It is a different product boundary. Agencies evaluating the bundled route can check Studio and Agency limits on the Trillet white-label pricing page.
For a page focused on replacing Vapi in an agency workflow rather than calculating its bill, see the Vapi alternative for agencies.
How Can Teams Reduce Vapi Cost per Minute?
The biggest controllable Vapi cost is usually the model input, followed by voice choice and telephony. Optimize the assistant before negotiating around a badly specified workload.
- Shorten the system prompt and tool definitions.
- Use prompt caching where the model supports it.
- Select the least expensive model that meets the task's accuracy requirements.
- Keep calls focused so conversation history does not grow without bound.
- Compare TTS on actual spoken-character volume, not a voice demo.
- Use web or negotiated telephony where it fits the user experience.
- Measure concurrency separately from total minutes.
- Inspect real call-level cost breakdowns before projecting high volume.
Vapi's component flexibility makes these optimizations possible. It also makes testing essential, because a quote based on a short demo call may understate a long production conversation.
Frequently Asked Questions
How much does Vapi cost per minute?
Vapi charges $0.05 per connected call minute for hosting without a Success Package, plus STT, LLM, TTS, and telephony. The named examples in this article total about $0.071, $0.093, and $0.215 per minute under low, mid-range, and premium assumptions.
Does Vapi's $0.05 include the LLM and voice?
No. The $0.05 is the Vapi platform and hosting fee. Model-provider costs for STT, LLM, and TTS are passed through at cost or billed through the customer's own API keys.
Does Vapi charge for web calls?
Yes. Web calls avoid the modeled PSTN carrier fee but still incur Vapi's $0.05 platform fee and the selected speech, model, and voice costs.
How much is Vapi HIPAA compliance?
Vapi lists HIPAA at $2,000 per month. A signed BAA and compliant STT, LLM, and TTS provider configuration are also required, and Vapi says HIPAA mode and ZDR cannot be enabled together.
Is Vapi cheaper than Trillet?
It can be. An optimized Vapi web stack can have a lower raw rate, while Trillet includes the white-label agency layer and bundles platform, STT, LLM, and TTS at $0.12 per AI minute. Compare the same call channel and the complete product scope.
Updated September 22, 2026: refreshed Vapi's public packaging, concurrency, retention, and Zero Data Retention details while preserving the component-cost scenarios.
