Vapi Pricing: The Real Per-Minute Cost Behind the $0.05 Rate

Vapi pricing starts at $0.05 per minute for the platform itself, and that number is the beginning of the bill, not the end of it. Vapi is voice infrastructure: the transcription, the language model, the voice, and the phone line are all separate services billed on top, either passed through at each provider's rates or paid directly if you bring your own API keys.

That control is why developers reach for Vapi. It is also why the bill is hard to predict before the whole stack exists. Published third-party breakdowns from CloudTalk, Telnyx, and Five put typical all-in costs between roughly $0.15 and $0.36 per minute once every layer is counted, three to seven times the advertised floor. Vapi's own numbers below were verified on vapi.ai/pricing on July 20, 2026 and can change at any time.

How Vapi pricing works#

Vapi sells the orchestration layer for voice agents. You still assemble the finished agent yourself. Its pay-as-you-go Build plan charges:

• $0.05 per minute of call time for hosting and orchestration
• $0.005 per message for SMS and chat
• 10 concurrent calls included, with additional concurrency lines at $10 per line per month

Model costs are passed through to you at provider rates. They are free from Vapi's side if you connect your own API keys and pay OpenAI, ElevenLabs, Deepgram, or your telephony carrier directly.

The Scale plan moves you to an annual contract: a fixed platform fee, committed usage volume with volume-based per-minute pricing, and the enterprise checklist (SOC 2, HIPAA, PCI, SSO, role-based access control, data residency, priority provider access, a support SLA, and a dedicated account team). Vapi does not publish Scale pricing; it is quoted.

The plans and their published add-ons compare like this before we break down each cost layer:

On Build, the base rate is the smallest part of the bill, and the two compliance add-ons stay fixed regardless of volume. A low-volume regulated deployment can spend more on HIPAA and Zero Data Retention than on its call minutes, so the plan you belong on depends as much on compliance needs as on usage.

The five layers of a Vapi call#

A production Vapi call assembles five meters. Only the first belongs to Vapi:

The spread between a cheap stack and a premium one is dramatic. A budget build (fast open models, standard voices) can stay near $0.10 all-in, while a frontier LLM with a premium cloned voice can push past $0.30 per minute. The same third-party guides linked above converge on $0.15 to $0.36 as the realistic production band, and community reports in Vapi cost discussions land in the same range.

Build vs. Scale: what each plan includes#

Two add-ons are priced publicly and apply to both plans: HIPAA compliance at $2,000 per month and Zero Data Retention at $1,000 per month. If you are in healthcare or handle regulated data, those two line items can exceed the entire usage bill of a mid-sized deployment.

The 14-day call history window on Build deserves more attention than it gets. If your operation needs recordings and transcripts for coaching, dispute resolution, or compliance retention, you either export everything continuously or move to Scale.

What a month actually costs#

Using the $0.15 to $0.36 all-in band from the published breakdowns above:

• 2,000 minutes per month: roughly $300 to $720
• 5,000 minutes: roughly $750 to $1,800
• 20,000 minutes: roughly $3,000 to $7,200, the volume where Scale contracts and committed-use discounts start to make sense

CloudTalk's guide estimates stable production deployments typically budget $40,000 to $70,000 per year once volume, engineering, and enterprise features are included. Whatever you assume per minute, add the cost that never shows up on the meter: Vapi takes technical capacity to assemble the stack, design the conversation, wire the integrations, and keep all five layers tuned as models and voices change. At 2,000 minutes a month the usage runs $300 to $720, but the engineering behind it is the bigger line, and for a revenue team without a developer to spare, that is the point where the math stops working.

Vapi pros and cons#

Every decision about Vapi comes back to the same tradeoff: control in exchange for the work of assembling and forecasting a stack.

Pros

• Full control over every layer. You pick the speech-to-text engine, language model, voice, and carrier, and swap any of them when a better or cheaper option appears.
• Bring-your-own-keys pricing. Connect your own OpenAI, ElevenLabs, Deepgram, or telephony accounts and the model layers cost $0 from Vapi's side, so you pay Vapi only the $0.05 orchestration fee.
• Low platform floor. At $0.05 per minute, the orchestration charge itself is among the cheapest in the category.
• Developer flexibility. Mid-call tool calls, provider swaps, and custom orchestration are all in scope for a team that can build them.

Cons

• Five meters to forecast. The advertised $0.05 hides four pass-through layers, which is why realistic all-in costs run $0.15 to $0.36 per minute instead of the headline rate.
• Engineering required. Assembling the stack, designing the conversation from scratch, wiring integrations, and tuning latency all stay on your team.
• Short retention on Build. Call history is kept 14 days, so coaching, dispute, or compliance recordings need continuous export or a Scale contract.
• Concurrency cap on Build. Ten simultaneous calls are included; each additional line is $10 per month.
• Compliance costs extra. HIPAA is a $2,000 per month add-on and Zero Data Retention is $1,000 per month, on top of usage.

The pros only pay off if someone on your team is going to use them. If you have engineers who want to own the model choice and negotiate provider rates, the control is worth paying for. Without them, the flexibility shows up as work you have to finish before a single call gets made.

Who Vapi fits#

Vapi is a strong choice for exactly one buyer: a team with developers that wants full control over every component of a voice agent. You pick the models, swap providers, orchestrate mid-call tool calls, and pay close to raw cost for each layer. When we compared nine calling platforms side by side, Vapi earned the maximum-control slot, with the caveat that both the cost forecasting and the conversation design stay on your plate.

The pricing model rewards that buyer too. Bring-your-own-keys means an engineering team that already has OpenAI and telephony relationships pays Vapi only the $0.05 orchestration fee and negotiates everything else at the source.

If that does not describe your team, the same architecture works against you: five meters to forecast, a conversation to engineer from a blank page, and latency and quality tuning that stays on your desk. Sales teams comparing the developer platforms directly should also read our Retell AI pricing breakdown, which bundles more of the stack into its published rate.

Where Rezora IO fits#

The number on the Vapi meter is the easy part of the budget. The rest is the engineering the meter never prices: the person who assembles the five layers, designs the conversation, and re-tunes it every time a model or a voice changes. On a do-it-yourself stack that cost is money you are already spending; it just never lands on an invoice.

Rezora IO is what you buy when you would rather not carry that line at all. It ships as a finished caller, so the assembly, the prompt design, and the ongoing tuning are already done and folded into one usage number. Its hosted models are fine-tuned on sales calls, which is why it can hold its own against a hesitant homeowner or a price objection without an engineer maintaining a script behind it.

The tradeoff is honest. You give up the layer-by-layer control Vapi sells, and in exchange you stop paying for the engineering seat that control requires. For an operator whose goal is leads getting called today, that missing line item is usually the whole reason to buy.

Pricing is usage-based, with enterprise plans, and it arrives as a single figure to reconcile each month instead of five separate meters.

The two approaches line up like this on the decisions that drive cost:

FAQ#

How much does Vapi cost per minute?#

Vapi's platform fee is $0.05 per minute of call time, prorated. The speech-to-text, language model, voice, and telephony layers bill on top at provider rates, which is why published guides put realistic all-in costs at roughly $0.15 to $0.36 per minute depending on the stack.

How much does Vapi cost per month?#

Build has no fixed monthly platform fee; you pay for what you use. At the $0.15 to $0.36 all-in band from published guides, 2,000 minutes runs roughly $300 to $720, 5,000 minutes roughly $750 to $1,800, and 20,000 minutes roughly $3,000 to $7,200 per month. Add-ons such as HIPAA ($2,000 per month) or Zero Data Retention ($1,000 per month) stack on top of usage.

Does the $0.05 per minute include the AI model?#

No. Model costs are passed through at the provider's rates, or billed directly by the provider if you connect your own API keys. The $0.05 covers Vapi's hosting and orchestration only.

What is included free on the Build plan?#

Build starts with 60+ minutes of included usage, 10 concurrent calls, SMS and chat at $0.005 per message, 14 days of call history, and community plus email support. Additional concurrency is $10 per line per month.

What are Vapi's concurrency and rate limits?#

The Build plan includes 10 concurrent calls, and each additional concurrency line costs $10 per month. Scale plans set custom concurrency as part of the contract. If you expect to run more than 10 simultaneous calls, budget for extra lines or move to Scale.

How much is Vapi's enterprise (Scale) plan?#

Vapi does not publish Scale pricing. It is an annual contract with a fixed platform fee, committed volume at discounted per-minute rates, and enterprise features like SOC 2, HIPAA, PCI, SSO, and a dedicated account team. HIPAA compliance and Zero Data Retention are published add-ons at $2,000 and $1,000 per month respectively.

Is Vapi HIPAA compliant?#

Vapi offers HIPAA compliance as a paid add-on at $2,000 per month on any plan, and it is part of the Scale plan's enterprise checklist alongside SOC 2, PCI, SSO, and RBAC. Standard Build usage does not include HIPAA by default. Healthcare and other regulated deployments should budget for the add-on or a Scale contract.

Is Vapi worth it?#

Vapi is worth it for a team with developers that wants full control over every component of a voice agent and can forecast five separate cost meters. That team pays Vapi only the $0.05 orchestration fee, brings its own provider keys, and comes out ahead. A revenue team that just needs leads called should skip it: you would carry the same $0.15 to $0.36 per minute plus an engineer to keep the stack running, for a caller you still have to design yourself. Buy Vapi if you are building a voice product; buy a finished caller if you only need leads on the phone.

Is Vapi cheaper than Retell AI?#

At the platform layer, Vapi's $0.05 per minute is below Retell's $0.07 to $0.08 voice engine rate, but Retell bundles text-to-speech into that rate while Vapi bills it separately. Once full stacks are assembled, both land in overlapping ranges; our Retell AI pricing guide has the component math to compare like for like.

What is better than Vapi?#

There is no single better tool; it depends on how much you want to build. Developer-focused alternatives like Retell AI bundle more of the stack into a published rate, as covered in our Retell AI pricing guide. If you want a finished caller instead of a kit, Rezora IO trains its own hosted models on real sales conversations so there is no stack to assemble.