How much does an AI voice agent cost per minute?
On list prices read November 12, 2026, a minute of AI voice agent time costs about 7 to 31 cents on a build-it-yourself platform, once the platform fee, telephony, speech, voice and language model are added together, and about 15 to 60 cents on a finished flat plan, depending on how many of the included minutes you use. The cheapest advertised number is rarely the whole price: a $0.05 platform fee is one of five or six charges. Botcadence's AI voice agent is $299 a month for 1,500 minutes, which is 19.9 cents a minute if you use them all.
Key takeaways
- A voice agent minute has up to six cost parts: platform, telephony, speech recognition, voice, language model and extras.
- Retell AI's own calculator shows $0.11 a minute for its default setup; Vapi's own estimate for 1,000 minutes is $82 to $129, which is 8 to 13 cents a minute; Bland AI charges $0.14 with the model included.
- Flat plans look expensive per minute when you use few minutes and cheap when you use most of them.
- Concurrency is a separate charge on some platforms: $8 a line a month on Retell above 20, $10 on Vapi.
- Ask whether the model, the carrier and the numbers are inside the quoted rate.
The six parts of a minute
| Part | What it is | List prices, read November 12, 2026 |
|---|
| Platform | The software that runs the call | Vapi $0.05 a minute; Retell AI $0.055; Twilio ConversationRelay $0.07 |
| Telephony | The phone network | Twilio: inbound local $0.0085, outbound $0.014 a minute. Telnyx: from $0.002 a minute plus a SIP fee |
| Speech recognition | Turning the caller's speech into text | Deepgram Nova-3 streaming $0.0077 a minute; AssemblyAI streaming $0.15 an hour, which is $0.0025 a minute |
| Voice | Turning the reply into speech | Billed per character: ElevenLabs Flash $0.04 per 1,000 characters; Deepgram Aura-2 $0.03. On Retell, $0.015 a minute for most voices and $0.04 to $0.10 for ElevenLabs |
| Language model | Deciding what to say | Billed per token. As a per-minute figure, Retell lists models from $0.0016 to $0.32 a minute, with mid-range models at about 1 to 5 cents |
| Extras | Numbers, concurrency, add-ons | Numbers $1 to $2 a month. Voicemail detection $0.0075 a call on Twilio. Knowledge base, noise reduction and similar add-ons at half a cent to a cent a minute on Retell |
Voice and model costs depend on how much is said. A talkative agent with a large model can cost several times a brief one with a small model.
| Platform | Its own all-in figure | What is included |
|---|
| Retell AI | $0.11 a minute in its calculator's default; $0.07 to $0.31 across options | Engine, voice, model; telephony $0.015 through Twilio or Telnyx |
| Vapi | $82 to $129 for 1,000 minutes, in its own estimate | $0.05 hosting plus model, speech and carrier at cost |
| Bland AI | $0.14 a minute on Start; $0.12 on Build plus $299 a month | Model, speech and voice; no token charges |
| ElevenLabs Agents | $0.08 for each extra minute | Platform and voice; the model and carrier are billed on top |
| PolyAI platform rate | From 9.5 cents a minute | Pay-as-you-go platform use |
These are the vendors' numbers, not measurements of ours. Each depends on the voice and model chosen.
Flat plans: the effective cost per minute
A flat plan's cost per minute depends on how much of it you use. The table divides the monthly bill by the minutes used.
| Plan | 500 minutes used | 1,000 minutes used | 1,500 minutes used |
|---|
| Botcadence, $299 for 1,500 | 59.8 cents | 29.9 cents | 19.9 cents |
| PolyAI, cheapest plan for the volume | 24.8 cents ($124) | 19.9 cents ($199) | 19.9 cents ($299) |
| Rosie, cheapest plan for the volume | 29.8 cents ($149) | 14.9 cents ($149) | 19.9 cents ($299) |
| Dialzara, cheapest plan for the volume | 39.8 cents ($199) | 34.9 cents ($349) | 34.9 cents ($524) |
| ElevenLabs Agents Pro tier, platform only | 19.8 cents ($99) | 9.9 cents ($99) | 8.0 cents ($120) |
ElevenLabs' figures exclude the language model and the carrier, which it bills separately. The flat plans from Botcadence, PolyAI, Rosie and Dialzara are finished products with the model included.
Two readings follow. At low usage a flat plan is the most expensive way to buy a minute. And a finished product at 20 cents and a platform at 11 cents are not the same purchase: the platform's price leaves out the person who builds and maintains the agent.
Concurrency: the charge people forget
Calls that happen at the same moment need concurrent capacity, and vendors treat it differently.
| Vendor | Concurrency terms |
|---|
| Retell AI | 20 lines included, then $8 a line a month |
| Vapi | 4 to 30 lines by plan, then $10 a line a month |
| ElevenLabs Agents | 4 to 40 lines by tier; burst up to three times the limit at double the per-minute rate |
| Bland AI | 10 lines on Start, 50 on Build |
| PolyAI | 5 or 10 lines by plan |
| Botcadence | Set per account; starts at one line and is raised on request |
A business that gets three calls in the same minute once a week needs three lines, not three times the minutes.
What changes the bill
- Call length. Every part except the number fee scales with minutes.
- How much the agent talks. Voice is billed per character on most providers.
- The model. Moving from a small to a large model can add 4 cents a minute or more.
- Billing increments. Per-second billing costs less than rounding each call up to a minute. Botcadence bills a 15-second minimum and then 30-second steps.
- Transfers. Some platforms charge a transfer rate; on Botcadence the agent's minutes stop once a person takes the call.
- Outbound. Outbound carrier minutes cost more than inbound, and voicemail detection is charged per call by some carriers.
- Failed and short calls. Ask whether calls that never connect, or last a few seconds, are billed.
How to estimate your own cost
- Take last month's minutes and your peak number of calls in progress together.
- For a platform: add the platform fee, your carrier's rate, and the platform's listed voice and model prices for the options you would choose. Multiply by minutes. Add numbers and extra lines.
- For a flat plan: take the plan that covers your minutes, add overage, and divide by the minutes you will really use.
- Add the cost of whoever builds and maintains the agent, if it is not the vendor.
- Repeat for your busiest month.
For vendor-by-vendor detail see Retell AI pricing explained, Bland AI pricing explained and the AI voice agent pricing guide. For the charges that surprise people, see AI voice agent billing: the fine print. The AI receptionist for small business page shows what a finished plan includes.