Skip to main content
Real Human Feedback

Benchmark profile · Voice AI platform

Vapi: pricing, features and what we test

Developer platform for voice agents with a free choice of speech and language models.
Not yet scoredTesting in progressVisit Vapi (opens the vendor's site in a new tab)

What it is and who it is for

A factual profile, not a review or a ranking.

Vapi is a voice AI platform for developers and agencies. It lets a builder choose the speech and language models behind an agent, run unlimited agents, and attach phone numbers, all through an API.

Vapi publishes a hosting fee of $0.05 a minute plus the cost of the models used, a Core plan at $29 a month and Pro from $999 a month.

It connects to Twilio and Telnyx for telephony and to OpenAI, Deepgram and ElevenLabs for models. The benchmark will score a named reference receptionist built on Vapi, disclosed on this page.

Published pricing

Pricing as published by the vendor, checked September 2026. Prices change, so confirm on the vendor's own pricing page before you buy.

Vapi pricing tiers as published
PlanPublished priceIncluded, as published
Hosting$0.05/minplus model costs
Core$29/moNot stated
ProFrom $999/moNot stated
  • The per minute hosting fee is published separately from the cost of the models used.

Checked against: vapi.ai/pricing (opens the vendor's site in a new tab)

Compare pricing across every product in the benchmark

Features

As listed by Vapi. Listing a feature is not the same as testing it.

  • Voice agent builder
  • Model choice
  • Unlimited agents
  • Phone numbers
  • API

Integrations

As published by the vendor.

  • Twilio
  • Telnyx
  • OpenAI
  • Deepgram
  • ElevenLabs

Industries targeted

The industries Vapi names on its own site.

  • Developers and agencies

Benchmark score

Scores appear only after real people place test calls.

Not yet scoredTesting in progress

No test calls have been made to Vapi yet, so there is no score, rating or ranking on this page. When real people have called a reference setup, the results will appear here against each of the nine rubric categories, with the setup disclosed. Because Vapi is a builder platform, the benchmark will score a named reference receptionist built on it, and name that receptionist here.

How the benchmark scores products

What our test calls will check

Every product is scored against the full nine category rubric. Given what Vapi publishes, these are the calls we will look at most closely.

  • 01Latency, silences and talk-over, which vary with the model choices in the build
  • 02Understanding accents and background noise, which depends on the speech model chosen
  • 03Invented information when a caller asks something outside the prompt
  • 04Handoff to a human over the connected telephony provider

Rubric categories: R1 Greeting and identification, R2 Understanding intent, R3 Booking accuracy, R4 Handoff to a human, R5 After hours and urgent calls, R6 Edge cases, R7 Invented information, R8 Latency and talk-over, R9 Message accuracy.

Already using Vapi?

The benchmark tests a reference setup. Yours is configured for your business, with your hours, your services and your callers, so it can behave very differently. Real people can call the Vapi receptionist you already have and send you a written report. Your results go to your company only.

Frequently asked questions

Short answers about Vapi.

How much does Vapi cost?

As published in September 2026: $0.05 per minute for hosting plus model costs, Core at $29 a month, and Pro from $999 a month.

Why does a Vapi receptionist sound different from one business to the next?

Vapi lets each builder choose the models behind the agent, so two receptionists on Vapi can behave very differently. That is why we test the receptionist you actually have.

Has Vapi been scored?

Not yet. It shows as not yet scored until real test calls have been placed to a disclosed reference receptionist.

Hear your AI receptionist the way your customers do

Tell us the number your AI receptionist answers and anything you want us to look at. Real people place the calls, and you receive a written report.