How AI Voice Agents Are Priced (and How to Compare Them Fairly)
Updated 2026-10-17. Prices come from vendor pages; see each vendor profile for sources.
Voice AI pricing pages are hard to compare because vendors charge for different things. One quotes $0.05 a minute, another $0.14, a third $49 a month. Those numbers can all describe the same call. Here are the five pricing models you'll run into, and how to turn each into one comparable number: your cost per month at your call volume.
1. Platform fee + pass-through models
The platform charges a per-minute hosting fee. Speech-to-text (STT), the language model (LLM), text-to-speech (TTS) and telephony are billed on top, usually at the provider's own rate. Vapi works this way: it lists $0.05/min for hosting, with models "passed through at cost". Its own estimator shows about $82–$129 for 1,000 minutes once models are included. LiveKit Cloud charges $0.01/min per agent session plus $0.01/min telephony, with model inference billed per minute on top.
How to compare: add up the stack. A typical stack from published prices (checked 2026-10-17):
- STT: about $0.0065/min (Deepgram Flux)
- TTS: from $0.015/min (Retell platform voices) to $0.04–$0.10/min (ElevenLabs voices on Retell)
- Telephony: $0.0085/min inbound, $0.014/min outbound (Twilio, US local)
- LLM: varies with model and prompt size, from under a cent to several cents per minute
2. All-in per-minute rate
Some vendors quote a single rate that includes the models. Bland's Start plan is $0.14/min with no platform fee; its Build plan is $0.12/min plus $299/month. Retell publishes a $0.07–$0.31/min range, depending on which LLM and voice you choose. Telnyx bundles orchestration, STT and TTS at $0.05/min and estimates about $0.06/min once LLM tokens and telephony are added. Always check whether telephony and transfer minutes are included. Bland, for example, charges transfer time separately.
3. Monthly plan with included minutes
Most AI receptionists for small businesses sell a monthly plan with a minute allowance and an overage rate. Dialzara starts at $29/mo for 60 minutes ($0.48/min overage) and goes up to $349/mo for 1,000 minutes ($0.35/min). Trillet is $49/mo for 150 minutes, then $0.20/min. Phonely's Starter is $50/mo for 200 minutes, with $0.25/min overage. When a vendor doesn't publish an overage rate (Rosie and My AI Front Desk, for example), you only know the cost up to the included minutes.
4. Per-call or per-caller pricing
Some receptionists charge per call or per unique caller instead of per minute. Smith.ai's AI Receptionist Pro is $150/mo, about $2.00 per call. Upfirst starts at $24.95/mo for 30 calls. Goodcall prices by unique customers served, starting at $79/mo. If your calls are long, per-call pricing can work out cheaper. If they're short, per-minute usually wins.
5. Enterprise contracts
Enterprise platforms often publish no prices, or only a floor. Synthflow says enterprise contracts start at $30,000 a year, and PolyAI, Cognigy and Regal are quote-only. For these, ask for a written quote that covers per-minute or per-conversation cost, concurrency, implementation fees and minimum commitment.
A fair comparison in four steps
- Estimate your monthly minutes: calls × average length. Inbound receptionist calls often run 1–3 minutes.
- For each vendor, work out the cheapest plan at that volume, including overage.
- Add model and phone-line costs wherever the vendor doesn't include them.
- Divide by minutes to get an effective cost per minute.
The cost calculator does these steps for every vendor with public plans.
Hidden costs to ask about
- Phone numbers (Twilio lists local numbers at $1.15/mo, for example)
- Transfer minutes and SMS
- Extra concurrency, extra seats or workspaces
- Compliance add-ons: Vapi lists a HIPAA/BAA package at $2,000/month
- Premium or cloned voices