Budgeting Vapi means separating call usage from monthly packages and peak capacity. For developers and product teams planning a voice agent, the useful budget shows which components run on every call, which fees recur even at low volume, and how shared call lines affect the deployment.
How much does Vapi cost?
Vapi's current public pricing sets Usage-only hosting at $0.05 per minute, with speech, model and telephony costs separate. An optional Success Package adds a recurring charge. Extra capacity and other add-ons can increase that total.
The monthly budget has three parts:
- Call usage: hosting plus the providers used by the agent.
- Package and add-ons: support, included limits and any extra capacity or data-handling options.
- Separately billed services: applicable carrier and other external charges, including phone-number charges only where provider or account terms require them.
A package does not replace usage charges. The hosting rate and package figures here describe Vapi's current public model. Account-specific or legacy arrangements should be checked in the Dashboard or agreement.
Vapi packages and included limits
A Success Package changes support and entitlements while usage billing continues. The Vapi pricing page separates the package charge from usage.
| Billing option | Monthly package charge, before usage |
|---|---|
| Usage only | No Success Package fee |
| Core | $29 |
| Pro | 10% of Vapi hosting fees, with a $999 minimum |
| Premier | Custom |
The Pro percentage applies to hosting spend. Provider costs remain separate. Its minimum also means the package fee does not fall to a few dollars in a low-volume month.
| Included entitlement | Usage only | Core | Pro |
|---|---|---|---|
| Concurrent call lines | 4 | 10 | 30 |
| Vapi phone numbers | 1 | 5 | 10 |
| Raw data retention | 14 days | 30 days | 180 days |
Premier limits depend on the agreement. An organization is a workspace; its allowance is distinct from the number of simultaneous calls. The table gives included Vapi phone-number counts. It does not price additional or externally purchased numbers; any charge for them depends on provider or account terms.
Support commitments differ too. Core includes a two-business-day response commitment. Pro adds a dedicated Slack channel and a one-business-day response commitment. A response commitment concerns support turnaround; an uptime commitment concerns service availability. They serve different requirements.
What goes into Vapi's per-minute cost?
A typical voice pipeline uses speech-to-text (STT) to transcribe audio, a large language model (LLM) to generate responses, and text-to-speech (TTS) to produce the voice. Hosting coordinates the running agent. Transport connects the call.
The Vapi calculator example uses Deepgram, OpenAI and ElevenLabs, with the Vapi Telephony / SIP transport entry. SIP stands for Session Initiation Protocol, a signaling protocol used in internet telephony.
Illustrative budget: Assume 1,000 monthly billable minutes, no Success Package and the displayed calculator estimates. The component amounts below multiply each displayed per-minute estimate by 1,000.
| Component | Calculator estimate per minute | Amount for 1,000 minutes |
|---|---|---|
| Vapi hosting | $0.0500 | $50.00 |
| Deepgram transcription | $0.0095–$0.0099 | $9.50–$9.90 |
| OpenAI model | $0.0077–$0.0452 | $7.70–$45.20 |
| ElevenLabs voice | $0.0146–$0.0238 | $14.60–$23.80 |
| Vapi Telephony / SIP transport | $0 | $0.00 |
| Component sum | $0.0818–$0.1289 | $81.80–$128.90 |
The calculator rounds that monthly estimate to $82–$129. This is a planning range for the selected components. It does not define a fixed rate for every Vapi call or name one exact OpenAI model.
The underlying meters are different
The calculator expresses costs per minute so they can be added together. Underneath, the components use different billing units:
- Transcription uses audio minutes. Vapi's estimate includes audio from both the caller and the assistant.
- The model uses input and output tokens. Prompt size, tool definitions, conversation history and caching affect the amount.
- The voice uses spoken-text characters. An assistant that speaks more generates more billable voice output.
Equal call durations can therefore produce different costs. A long instruction prompt or large tool definitions can increase model spend without adding a minute to the conversation. Longer calls can also accumulate history that gets sent to the model again.
The useful distinction is between a normalized calculator estimate and the quantities a particular assistant consumes. Treating every provider as a fixed call-minute meter hides that difference.
Transport can change the sum
The selected $0 transport entry leaves hosting and AI usage charges intact. It also does not make an independent carrier contract free.
Vapi's transport price table lists Twilio inbound at $0.008 per minute and outbound at $0.014 per minute, charged by the provider. For 1,000 matching transport minutes, those entries add $8 or $14 respectively. Using the same AI components, that produces an illustrative Usage-only total of $89.80–$136.90 inbound or $95.80–$142.90 outbound.
Those examples apply the listed direction-specific rates. Different destinations, number types, carrier terms or billable quantities can change the external charge.
Monthly examples: adding a package and call lines
Package charges sit on top of the component estimate. Keeping the same 1,000-minute workload makes their effect visible without changing the model, voice or transport at the same time.
Assume a full month of each recurring charge, with the same component estimate above. For the capacity example, assume Core plus two additional call lines, each priced at $10 per month.
| Configuration | Illustrative monthly sum |
|---|---|
| Usage only | $81.80–$128.90 |
| Core | $110.80–$157.90 |
| Core plus two extra lines | $130.80–$177.90 |
| Pro, applying its displayed minimum | $1,080.80–$1,127.90 |
The Core calculation adds $29. Two extra lines add another $20, giving a $49 recurring package-and-capacity charge before usage.
The Pro row applies the stated monthly minimum: hosting for this workload is $50, and 10% of that is $5. The $999 floor controls the illustration. Usage remains additional.
These sums exclude taxes and other optional services. No charge for additional or external phone numbers is assumed: Vapi's displayed table does not price them, and any charge depends on provider or account terms.
The examples assume the displayed component estimates scale across the stated minutes. They do not infer a call-rounding rule, failed-call treatment or a negotiated discount. A purchase partway through a billing period can also have a different amount due because of its effective date and proration.
A small monthly minute total can coexist with a high peak call count. That is why the usage estimate and capacity charge belong on separate budget lines.
Concurrency is shared across organizations
Concurrency means calls active at the same time. One active call occupies one call line. Under the current model, organizations within a subscription share its capacity. When the pool is full, Vapi cannot start another outbound call or accept another inbound call until capacity becomes available. These are subscription-level limits.
For example, a Core subscription containing two organizations has ten shared lines. If one organization occupies seven, three remain for the other. Creating another organization does not create another ten-line pool.
The Core example with two extra lines has a twelve-line entitlement. It is a capacity allowance, not a guarantee that twelve lines are sufficient for any traffic pattern. Inbound traffic and outbound batches compete for the same pool. A campaign's maximum-concurrency setting limits that campaign; it does not purchase or reserve additional subscription capacity.
This matters for a product serving several customers. Monthly minutes describe total consumption. The overlap between their busy periods determines the simultaneous capacity needed. Two deployments with equal monthly minutes can need different line counts.
Add-ons and other charges to budget separately
Capacity add-ons can accompany Usage only or a Success Package. A separate Health Insurance Portability and Accountability Act (HIPAA) data-handling add-on requires a signed Business Associate Agreement and compatible providers. The displayed add-on prices are:
| Add-on | Displayed monthly price |
|---|---|
| Additional concurrent call line | $10 per line |
| Additional organization | $20 per organization |
| HIPAA data-handling add-on | $2,000 |
Purchasing the HIPAA add-on alone does not establish that a complete deployment meets every applicable requirement. Vapi's add-on requirements also distinguish organization-specific data handling from shared call capacity.
Some usage categories are separate from the basic speech pipeline. A call's cost breakdown can include analysis, voicemail detection and knowledge-base usage. Their presence depends on the call and enabled services; there is no single fixed amount to add to every call. Cost visibility can also differ for invoiced accounts.
Prepaid credits fund usage
Usage only has no recurring Success Package fee, and new accounts start with $5 in usage credits. Ongoing calls still consume paid usage.
Under Vapi's prepaid credit system, the subscription balance funds hosting and provider usage. If it reaches zero with auto-reload off, the subscription freezes until more credits are purchased. A recurring package and a funded usage balance serve separate purposes: the first buys entitlements, while the second funds calls.
For a wider budget that includes engineering and operating work across platforms, our voice AI pricing guide separates those costs from the advertised minute rate.
Where Dasha fits a production-platform decision
Price is one input to choosing the runtime behind a conversational product. We build Dasha for technical teams that want a managed production runtime with dashboard and REST API workflows. Our dashboard supports agent configuration, browser testing and completed-call inspection. Custom integrations and production acceptance still require engineering work.
Our Vapi alternatives guide covers the broader platform-choice tradeoffs.
To evaluate our managed runtime with your own call flow, get started with Dasha.
