Vapi Pricing Explained: Platform Fee, Models, Phone and Add-Ons

Vapi's bill has more parts than its $0.05-a-minute hosting fee. Speech-to-text, the AI model, the voice and the phone line are billed on top at each provider's price. An optional plan (Core $29, Pro from $999 a month) or HIPAA at $2,000 a month can add more, so your total depends on your models and minutes.

Key Takeaways

• Vapi's $0.05 a minute is only its own fee; the AI and phone line cost extra.
• If you make few calls, the Pro plan's $999 minimum can be most of your bill.
• Vapi's calculator uses a discounted speech-to-text price, so budget on regular prices.

This post is part of our guide to Vapi alternatives. The guide compares the options, and this post adds up what staying on Vapi costs. Every price below is Vapi's own list price, and every sum is shown so you can check it.

The five parts of a Vapi bill

Every Vapi minute has five charges, and only one of them goes to Vapi.

The first is hosting. Vapi charges a prepaid $0.05 a minute. If your credit balance reaches $0 and auto-reload is off, no new calls can start.

The other four lines are speech-to-text, the language model, the voice and the phone line. Vapi bills each at the provider's listed price "with no markup". With your own provider key, that provider bills you directly, and the $0.05 hosting fee still applies.

One detail changes the math. Vapi transcribes both the caller and the assistant, so speech-to-text is billed on two audio channels. Vapi's guide to how it estimates model cost also warns that long calls usually cost more than the estimate, because the growing conversation is sent again with each request.

This is Vapi's current plan, introduced in mid-September 2026. If you joined earlier, your bill may still follow older terms.

Vapi's four plans side by side

Vapi's plans, which it calls Success Packages, set support, call limits and data retention. They do not change the $0.05 hosting rate.

Add-ons work on any plan, including usage only. HIPAA handling costs $2,000 a month and comes with a business associate agreement (BAA). Extra concurrent call lines cost $10 a line per month, and extra organizations $20 each per month.

Concurrency counts across the whole subscription. When every line is busy, you cannot start an outbound call or answer an inbound one until a line frees up.

The free Vapi numbers are US numbers for inbound calls only, and they cannot be ported out. Outbound calls need a number you import, for example from Twilio. If you are choosing between two hosted platforms, our Vapi vs Retell AI comparison puts these plans next to Retell's.

When the Pro package is most of the bill

At low volume, Pro's monthly minimum is bigger than everything else on the bill combined.

We ran Vapi's own calculator at 1,000 minutes a month, with its default Deepgram, OpenAI and ElevenLabs settings, Vapi telephony and Pro selected. It returned $50 for hosting, $10 for transcription, $8 to $45 for the model, $15 to $24 for the voice, $0 for the phone line and $999 for Pro. The estimated total was $1,081 to $1,128 a month. With no package, the same inputs came to $82 to $129.

So Pro turns roughly 8 to 13 cents a minute into roughly $1.08 to $1.13 a minute. The 10% rule only matters at scale. Ten percent of $0.05 is half a cent a minute, so the $999 minimum holds until about 199,800 minutes a month. Even at 1,000 minutes a day, 10% of hosting is about $150 a month, and Pro still adds about $12,000 a year.

Pro makes sense when you need its 99% uptime SLA and faster support. If all you need is 30 lines, Core plus 20 extra lines gets you there for $229 a month ($29 + 20 x $10).

Open Source Alternative to Vapi / Retell

Self-hosted voice agent platform — no per-minute fees

dograh-hq/dograh

Star on GitHub

What Vapi costs for three example businesses

Vapi has no single price. The bill depends on the services you pair with it: a speech-to-text model, an AI model, a voice, a phone line and, if you want one, a Vapi plan. Below are three example businesses, each with its own mix and every cost added up.

Every price is the regular list price on the provider's own page. Deepgram and ElevenLabs both run promotions right now, and we ignore them, because a promotion is a poor base for a yearly budget. That is why our speech-to-text line is higher than Vapi's calculator, which uses Deepgram's promotional $0.0048 a minute.

Here are our assumptions, so you can redo the math with your own numbers. Speech-to-text is Deepgram Nova-3 at $0.0077 a minute, billed on both channels, so $0.0154. For the model we use Vapi's own estimate for its default OpenAI model, $0.0077 to $0.0452 a minute, because the real figure depends on your prompt length. For reference, gpt-5-mini costs $0.25 per million input tokens and $2.00 per million output tokens.

For the voice, we assume the agent speaks for about half of each call minute, which we put at 400 characters of text. Cartesia counts 750 to 800 characters for a full minute of audio. Taxes, call recording, extra numbers and volume discounts are left out.

A small business answering calls

This is a small support line: 1,000 minutes a month, no package, and the free Vapi number.

At 1,000 minutes, that is $89.10 to $126.60 a month, with 4 concurrent calls.

A team making outbound calls

This is a reminder or sales line: 10,000 minutes a month of outbound calls on a Twilio number, with the Core package and a Cartesia voice.

That is $871 to $1,246 for the minutes. Twilio's rate comes from Twilio's US voice pricing.

Cartesia sells voice by monthly plan. Its Startup plan is $49 a month for 1.25 million characters. At 400 characters a minute, 10,000 minutes need 4 million, so 2.75 million go over the plan at $45 per million (overage must be switched on), which is $123.75.

Add the monthly items: Core at $29 for 10 lines, Cartesia at $172.75 ($49 + $123.75) and one Twilio local number at $1.15, for $202.90. The month comes to $1,073.90 to $1,448.90, or about 11 to 14.5 cents a minute.

A clinic taking patient calls under HIPAA

This is a clinic line: about 1,000 inbound minutes a day (roughly 30,000 a month) on Twilio, with Pro and the HIPAA add-on.

That is $2,928 to $4,053 for the minutes. Vapi quotes Twilio inbound at $0.008 a minute, while Twilio itself charges $0.0085, so we use Twilio's own rate.

Then the monthly items. Ten percent of $1,500 in hosting is $150, so Pro's $999 minimum applies. HIPAA adds $2,000, and the Twilio number $1.15, for $3,000.15. The month comes to $5,928.15 to $7,053.15, or about 20 to 23.5 cents a minute. Over a year, that is about $18k in hosting fees alone, before models and phone ($0.05 x 365,000 minutes = $18,250).

Two cautions apply to this setup. Vapi's HIPAA add-on needs compatible providers, so check that each provider you use offers its own HIPAA terms before you go live. HIPAA mode also cannot be combined with zero data retention. A BAA hands obligations to the vendor, but the clinic stays responsible for its own compliance.

These are three examples, not Vapi's price. Swap in your own models, call lengths and volume, and the same lines give you your number.

What changes with open-source Dograh

Moving to Dograh changes the hosting line, while the model and phone lines stay whatever providers you choose.

Self-hosted open-source Dograh has no per-minute platform fee and no package minimum. You pay for the server it runs on, at least 8 GB of RAM and 4 vCPUs with no GPU when you use model APIs, and you keep paying your model and phone providers. Dograh Cloud, the managed version, charges a 1¢ a minute platform fee. For the outbound team's 10,000 minutes, the platform line is $500 a month on Vapi and $100 on Dograh Cloud.

The healthcare line works differently. Dograh Cloud publishes no BAA, so for HIPAA calls the path is self-hosting. That removes Vapi's hosting, Pro and HIPAA lines from the clinic example, about $4,500 a month ($1,500 + $999 + $2,000), and replaces them with servers and the people who run them. Each model and phone provider that handles patient data still needs its own BAA, unless you run open-weight models on your own servers.

Self-hosting is real work. You run the server and apply upgrades yourself, with community support and no vendor SLA unless you buy Enterprise. Dograh can also shrink the voice line by playing pre-recorded clips for fixed lines, as our guide on how to cut voice AI costs by mixing pre-recorded audio with TTS shows.

For every row beyond cost, read Dograh vs Vapi, and for licences and self-hosting, our open-source alternative to Vapi guide. The code is on Dograh's GitHub under BSD-2-Clause, so you can price your own setup before you move a single number.