What AI voice-agent platforms charge
AI voice-agent platforms publish per-minute rates from about $0.05 to $0.31, and many of those rates leave out the language model, transcription, or telephony, which bill as separate line items. Callable’s own default stack runs about $0.062 a minute, itemized by carrier, transcription, model, and voice, plus $1.15 a month for one US phone number.
The comparison below records exactly what each platform states on its own pricing page, read on 20 September 2026. Where a platform does not publish a figure, the table says so instead of estimating one.
How the platforms compare
Prices as published on 20 September 2026. The table itemizes the per-minute rate, what bills separately on top of it, any plan minimum, concurrency charges, and monthly phone number rental, with a link to the page each figure was read from.
Prices as published on 20 September 2026.
| Platform | Per-minute rate | Billed separately | Plan minimum | Concurrency | Number rental | Source |
|---|---|---|---|---|---|---|
| Callable | About $0.062/min, itemized by layer | The per-minute rate is the sum of four layers: carrier, transcription, model, and voice. | No plan price is published. Usage is billed per call. | Not published | $1.15/month for one US number | See the price book |
| Vapi | $0.05/min platform hosting on the free-start and Core plans; Pro prices hosting at 10% of that fee | Transcription $0.0095 to $0.0099/min, model $0.0077 to $0.0452/min, voice $0.0146 to $0.0238/min, telephony free on Vapi’s own carrier or $0.0055 to $0.014/min via Twilio, Vonage or Telnyx | Free to start; Core $29/month; Pro $999/month minimum | 10 concurrent calls on Core, 30 on Pro; extra lines $10/month each | Not published as a standalone rate. Numbers are bundled into the plan. | vapi.ai/pricing |
| Retell AI | $0.07 to $0.31/min for voice agents, pay as you go | Voice infrastructure $0.055/min, text to speech $0.015 to $0.040/min, the language model priced per model, telephony free with your own SIP trunk or from $0.015/min through Retell | None published. Pay as you go, with $10 in free credit. | 20 concurrent calls included, then $8/month per extra concurrent call | $2/month per phone number ($10/month for a verified number) | retellai.com/pricing |
| Bland AI | $0.14/min talk time on Start, $0.12/min on Build | Transfer time only ($0.05/min Start, $0.04/min Build). The talk rate already includes the model, transcription, and voice. | Start has no platform fee; Build is $299/month | 10 concurrent calls on Start, 50 on Build | Not published as a standalone rate. Start includes one inbound number, stated as a $15/month value. | bland.ai/pricing |
| Synthflow | Not published | Not itemized on the pricing page | Enterprise contracts starting at $30,000 a year | Not published | Not published | synthflow.ai/pricing |
| ElevenLabs | $0.08/min beyond the plan’s included minutes; $0.16/min in burst mode above the concurrency limit | The language model and any telephony, billed separately based on usage, per ElevenLabs’ own pricing page | Free plan includes 15 minutes; paid plans run $6 to $990/month with 75 to 12,375 minutes included | 4 to 40 concurrent calls depending on plan | Not published on this page | elevenlabs.io/pricing/agents |
What a per-minute price usually leaves out
A single per-minute number rarely covers everything a call touches. Four layers usually bill separately: the language model that decides what to say, the speech-to-text engine that transcribes the caller, the text-to-speech voice that speaks the reply, and the telephony carrier that connects the call. Some platforms bundle two or three of these into one blended rate; others price each one on its own line.
Two more costs are easy to miss in a headline rate. Concurrency is how many calls a plan can run at the same time; going over it usually adds a monthly charge per extra line, not a per-minute one. Phone number rental is a flat monthly fee charged whether a number takes ten calls or none, so it never shows up in a per-minute figure at all.
A single number, on its own, does not describe what a call will actually cost until these extras are added back in.
How to compare two quotes fairly
Two per-minute numbers are only comparable once they are priced on the same footing. Start with the same total minutes on both sides, since included minutes and an overage rate change the effective per-minute cost as volume grows.
Next, check whether the quoted rate includes the model, the transcription, and the voice, or whether one or more of those bill separately. A lower headline rate that excludes the model is not necessarily cheaper than a higher one that includes it.
Finally, add anything billed monthly rather than per minute: phone number rental, a platform fee, or a charge for concurrency beyond what a plan includes. Only the resulting monthly total, at the same call volume, is a fair comparison.
A worked example: 300 calls of 3 minutes
300 calls a month at 3 minutes each is 900 minutes. On Callable’s default stack, at about $0.0619 a minute across carrier, transcription, model, and voice, that is about $55.71 in call minutes plus $1.15 for one US phone number, for about $56.86 a month, or about $0.19 for a single 3-minute call.
Run this same scenario, or substitute different numbers, in the calculator linked below.
Sources
Every competitor figure above is read from the platform’s own pricing page, linked here with the date it was checked.
Straight answers.
Because they price different things. A rate that only covers platform hosting looks cheaper than one that bundles the model, transcription, and voice, even when the all-in cost ends up similar.
No. A lower rate that excludes the model or the voice can cost more once those are added back in, especially at higher call volumes where plan minimums and concurrency charges start to matter.
Concurrency is how many calls a plan can run at the same time. Most platforms cap it in the base plan and charge a flat monthly fee per line beyond that cap, separate from the per-minute rate.
Usually not. A phone number is typically billed as a flat monthly fee, whether it carries one call or a thousand, so it sits outside any per-minute figure.
Often enough that a rate should be checked on the platform’s own page before it is used in a quote. This comparison states the date each figure was read and links to the source.
See /cost/calculator, which itemizes Callable’s default stack by carrier, transcription, model, and voice for any number of calls, minutes, and phone numbers.