Vapi Review 2026: Pricing, Features and Verdict

What it is

Vapi is an API first platform for voice agents. It orchestrates a phone or web call between a speech to text provider, an LLM and a text to speech voice, and lets you choose each one. The homepage lists providers including OpenAI, Anthropic, Gemini, Groq and Perplexity for the LLM layer, and Deepgram and ElevenLabs among the speech vendors. Vapi says more than 1M developers use it.

You configure assistants, tools and phone numbers through the API or dashboard. It is closer to infrastructure than a finished product: the flexibility is the point, and so is the expectation that someone on your team writes code.

Pricing explained

Vapi bills a prepaid hosting fee per minute, with providers passed through at cost. Here are the published ranges from its pricing page in October 2026.

Component	Price
Vapi hosting	$0.05 per minute
Transcription (Deepgram)	$0.0095 to $0.0099 per minute
LLM (OpenAI)	$0.0077 to $0.0452 per minute
Voice (ElevenLabs)	$0.0146 to $0.0238 per minute
Vapi telephony, SIP, WebSockets, Daily WebRTC	Free
Twilio inbound	$0.008 per minute
Twilio outbound	$0.014 per minute

Package	Price	Concurrent calls	Phone numbers	Retention	Support
No package	$0 to start, $5 credits	4	1	14 days	Discord community
Core	$29 per month	10	5	30 days	Email, 2 business day SLA
Pro	$999 per month minimum	30	10	180 days	Email plus Slack, 1 business day SLA
Premier	Custom	Elevated	Custom	Elevated	Named team, 1 hour P0 SLA

Add ons: HIPAA $2,000 per month, extra concurrency $10 per line per month, extra organizations $20 per month.

All in math. Low end on Vapi telephony: $0.05 + $0.0095 + $0.0077 + $0.0146 = $0.0818 per minute. High end of every range on Vapi telephony: $0.1289. High end with Twilio outbound: $0.1429.

Monthly bills. At 5,000 minutes, $409 low or $715 high, plus $29 if you need Core's 10 lines. At 20,000 minutes, $1,636 low or $2,858 high. If that 20,000 minutes runs through a busy line needing 30 calls at once, compare $999 Pro against Core plus 20 extra lines at $10 each ($229 per month). Core plus lines is cheaper unless you need the Slack support and retention.

The HIPAA math. $2,000 per month divided across 5,000 minutes adds $0.40 per minute. Across 100,000 minutes it adds $0.02. Vapi's HIPAA pricing only makes sense at very high volume; Retell and Telnyx include HIPAA.

See how Vapi compares on the AI Price Index and in the cost per minute breakdown.

Features

• Provider choice. Swap transcription, LLM and voice vendors per assistant.
• Latency. Vapi states under 500 ms average.
• Telephony. Free Vapi numbers and SIP, Twilio, WebSockets and Daily WebRTC for web calls.
• Languages. Set by the providers: Deepgram 100+, Google speech to text 125+, Gladia 110+, Azure voices 140+, ElevenLabs 30+, OpenAI voices 50+. Several transcribers detect language automatically mid call.
• Compliance. SOC 2, HIPAA and PCI listed for enterprise, with HIPAA as a paid add on.
• Organizations. 2 on the free and Core tiers, 10 on Pro, useful for agencies separating clients.

Setup

A developer can get a test assistant on a Vapi number quickly, but production work is code: you define the assistant, write tool endpoints, handle webhooks and tune provider settings for latency. There is no visual flow canvas comparable to Retell's. Expect free tier help to come from the Discord community rather than a support desk, and budget for Core if you want email support.

Who it is for

• Startups building a voice product they will resell, where margin per minute matters.
• Engineering teams that want to mix providers, test new models quickly and avoid lock in.
• High volume projects where pass through pricing beats a bundled rate.

Who should skip it

• Non developers. Retell AI or ElevenLabs Agents give you a visual builder.
• Small healthcare teams. The $2,000 per month HIPAA add on outweighs any per minute savings.
• Anyone needing more than 4 lines on day one without paying. Retell includes 20 and Telnyx 500.
• Buyers who want one simple rate. Bland bundles the models into one connected minute price.

Reviews

Vapi's Trustpilot profile is unclaimed and shows a 2.4 score from 15 reviews, with 67% one star. Complaints mention 3 to 5 second pauses, features that do not behave as documented, thin human support and charges beyond the platform fee. Fifteen reviews is a small sample for a platform reporting 1M+ developers, but the themes match the tradeoffs of an infrastructure product: configuration matters, and support costs extra.

Vapi vs the alternatives

Need	Vapi	Better option
Lowest platform fee with models at cost	$0.05 hosting, pass through	Vapi
Lowest complete rate	About $0.082	Telnyx, about $0.06 per its pricing page
Visual builder	No canvas	Retell AI
Free concurrency	4 calls	Retell AI (20), Telnyx (500)
HIPAA	$2,000 per month	Retell AI or Telnyx, included
Best voices	ElevenLabs pass through	ElevenLabs Agents, 5,000+ voices

The pattern is clear: Vapi wins when control over providers is the priority and loses when the buyer wants defaults included. If you are evaluating it against Retell, the deciding question is usually whether you will pay for Core or Pro anyway. If yes, Vapi's lower minute rate starts to count. If you would otherwise stay on Retell's free tier with 20 lines and HIPAA, Retell is cheaper in practice.

Verdict

Vapi is the right tool when an engineering team owns the agent and wants control and pass through pricing. For everyone else, the low headline rate is offset by paid concurrency, paid support and paid HIPAA. Compare it line by line in Retell AI vs Vapi, or see the full best AI voice agents ranking.