Cost and ROI · 6 min read

What Per-Minute Pricing Hides in AI Call Quotes

We priced 300, 1,000 and 3,000 calls a month on Retell, Bland and Vapi published rates. Platform fees can more than triple the headline per minute rate.

A Las Vegas dealership service department that loses calls during the Saturday rush usually gets the same pitch from AI phone vendors: one price per minute. The number looks small, and it's the least useful figure on the quote. What you pay each month depends on how many calls you take, how long they run, which AI model does the talking, and which flat fees come with the rate.

We read the published pricing pages of three platforms that sell voice agents by the minute: Retell AI, Bland AI, and Vapi. Then we built a monthly bill at three call volumes from their posted rates alone. Every vendor figure below was read on 2026-09-06. The math is ours. The one input no vendor publishes is how long your calls last, so we assumed a number and labeled it as an assumption.

The published rates

The Retell AI pricing page splits its price into parts. Voice infrastructure is $0.055 a minute, text-to-speech is $0.015 (ElevenLabs voices $0.040), and telephony is $0.015. The LLM line depends on the model: $0.025 for Claude 4.5 Haiku, $0.045 for GPT 4.1, $0.08 for Claude 4.6 Sonnet, and $0.16 for GPT 5.5. Phone numbers are $2 a month and there is no annual contract. Retell states its own all-in range as $0.07 to $0.31 a minute.

The Bland AI pricing page bundles the model, speech, and voice into one talk-time rate and says it passes through no token or provider charges. The Start plan is $0.14 a minute with no platform fee, capped at 100 calls a day and 10 at once. Build is $0.12 a minute plus a $299 monthly platform fee. Scale is $0.11 plus $499. Transfer time, meaning the minutes a call spends after being handed to a person, is billed on top at $0.05, $0.04, and $0.03 a minute on those three plans.

The Vapi pricing page charges a platform fee of $0.05 a minute, then bills speech-to-text, the LLM, and text-to-speech at each provider's cost, or $0 if you bring your own API keys. Those provider costs aren't on Vapi's page, so we can only show Vapi's platform fee. Treat it as a floor, not a total.

The assumption

We use a 3.5 minute average handle time for every call. It's a round placeholder for a short booking or intake call, not a measured number, and every total below moves in direct proportion to it. That gives 1,050 minutes at 300 calls a month, 3,500 minutes at 1,000 calls, and 10,500 minutes at 3,000 calls.

For Retell we priced two builds from its components. Lean: $0.055 infrastructure + $0.015 voice + $0.025 Haiku + $0.015 telephony = $0.11 a minute. Premium: $0.055 + $0.040 ElevenLabs + $0.08 Sonnet + $0.015 = $0.19 a minute. Both add one $2 number, and both fall inside Retell's stated range.

A month of calls, priced

Calls per monthMinutesRetell leanRetell premiumBland StartBland BuildVapi platform fee only
3001,050$117.50$201.50$147$425$52.50 + providers
1,0003,500$387$667$490$719$175 + providers
3,00010,500$1,157$1,997$1,470$1,559$525 + providers

Each cell is minutes times the rate, plus flat fees. Retell lean at 300 calls is 1,050 × $0.11 = $115.50, plus $2 for the number. Bland Build at 300 calls is 1,050 × $0.12 = $126, plus $299. The table leaves out transfer minutes, extra concurrent lines, any Bland number fees (they weren't among the rates we recorded), and tax.

One cell needs a warning. Over a 30 day month, 3,000 calls averages exactly 100 a day, which is Bland Start's daily cap. Any busy Saturday would push past it, so the $1,470 figure is for a plan you couldn't actually stay on.

What the per-minute number hides

Flat fees flip the ranking at low volume. Bland Build has the lower rate, $0.12 against Start's $0.14. At 300 calls it still costs $425 against $147, because the $299 fee is spread over only 1,050 minutes. That works out to $425 / 1,050 = about $0.40 per minute, more than triple the quoted rate. The two plans only break even at $299 / $0.02 = 14,950 minutes a month. That's about 4,270 calls at our handle time, which Start's daily cap wouldn't let you reach anyway. You buy Build for its headroom of 2,000 calls a day and 50 at once, not for its rate.

The model line is the biggest swing within one vendor. Retell lean and premium handle the same calls for the same minutes, and premium costs 73% more ($0.19 / $0.11 = 1.73). At 1,000 calls that's $667 minus $387, or $280 a month. A quote that says "from $0.07" tells you nothing until it names the model and the voice.

Handle time multiplies everything. If your real average is 5 minutes, minutes rise 43% (5 / 3.5 = 1.43). At 1,000 calls, Retell lean goes from $387 to 5,000 × $0.11 + $2 = $552. Bland Build goes from $719 to 5,000 × $0.12 + $299 = $899.

Add-ons are billed by the minute, the call, or the month. Retell charges $0.005 a minute for a knowledge base and $0.01 a minute for PII removal, and branded caller ID is $0.10 per outbound call. If you call back 300 missed callers a month with branded ID, that's $30 before any talk time. Retell includes 20 concurrent lines and charges $8 per extra line a month. Vapi's Build plan includes 10 concurrent calls at $10 per extra line. Vapi's HIPAA add-on is $2,000 a month and Zero Data Retention is $1,000 a month, and either one costs more than every Vapi platform fee in the table. Vapi's Build plan also includes 60 or more minutes, which covers under 6% of even the 300 call month.

Transfers are billed time too. On Bland, the minutes a call stays connected after a handoff are charged at the transfer rate. As an illustration, assume 20% of 1,000 calls transfer and each spends 2 minutes in transfer. That's 400 minutes, or $20 on Start and $16 on Build. For a service lane that's small. It grows for a business like a personal injury firm that wants the agent to screen callers and then hand most of them to a person.

Where this fits

Per-minute platforms are only one way to buy phone automation. Our breakdown of AI receptionist cost covers the wider range of options. The rates above also only cover the call itself. If the agent has to look up a repair order or a case status, connecting it to your systems is separate work, the kind of build we describe under custom AI assistants.

This model has limits. Prices change, and these were read on one day. Vapi's provider costs aren't modeled. We didn't test call quality, latency, or how often each agent needs a human.

What to do this week

  1. Pull 30 days of call logs from your phone system or carrier. Write down total inbound calls, missed calls, and average call length, and use your numbers in place of our 300 to 3,000 calls and 3.5 minutes.
  2. Ask every vendor for a written monthly total at your volume that lists the model, voice, platform fee, number fees, transfer billing, and concurrency limit. If a quote gives only a per-minute rate, rebuild it with the method in the table.
  3. Decide whether the agent answers every call or only the ones your team misses. If you price only missed and after-hours calls, your minute count, and your bill, covers just that slice of your volume.

If you'd like a second look at a quote, our free 15-minute audit runs your call numbers against a missed call recovery setup.

Drafted with AI assistance, researched, edited, and fact-checked by Elias Musleh on September 10, 2026.

More in Cost and ROI

/ FREE 15-MIN AUTOMATION AUDIT

Want this handled for you?

We build the automation, run it, and hand you the numbers. Book a fifteen minute call and we will tell you straight whether it is worth doing for your business.

Book your free audit

Prefer to skip the form? vendors@vegasbusinessai.com /702.773.8839