How much does an AI phone agent cost? The five pricing models, explained

The honest answer is a range: for a business handling 2,000 five-minute calls a month, published pricing across major platforms runs from roughly $200 to over $3,000 per month — and the spread comes less from vendor generosity than from five genuinely different pricing models. Some vendors sell a complete per-minute price; some an infrastructure line item that grows once a model and voice are attached; some a hosting fee that excludes the providers doing most of the work; some bundled tiers with platform fees and caps; and some skip minutes entirely and charge per agent or per call.

A buyer who compares headline numbers across those models will rank vendors incorrectly. This guide explains each model with a verified example, then runs one explicit workload through all five. Every third-party figure links to the vendor's own page with the date it was checked; all arithmetic is illustrative usage math, not a quote.

The five pricing models

1. All-in per-minute (example: ThunderPhone)

The simplest model: one per-minute rate that covers the complete engine — speech recognition, language model, and voice. ThunderPhone prices this way, pay as you go with no subscription or seat fee: Spark at 2¢/minute, Bolt at 5¢/minute, and Storm at 9¢/minute. Premium voice or language paths can add up to 3¢/minute, and optional add-ons carry published per-minute surcharges (verbal acknowledgements +2¢ and call supervision +8¢ on Storm, long prompts +1–2¢ per additional 10K tokens after the first 5K, demo numbers +1¢/minute plus $1/month). Hold bills a flat 2¢/minute; voicemail and phone-menu minutes bill at the engine rate, with straight-to-voicemail calls capped at one minute. The builder displays the current all-in rate; telephony depends on setup. (ThunderPhone published rates, 2026-08-19.)

Here the headline number is the usage price: multiply minutes by the rate and you have the bill.

2. Component-based (example: Retell AI)

Here each part of the call is its own line item. Retell lists voice infrastructure at $0.055/minute, language models from $0.003 to $0.16/minute by model, text-to-speech at $0.015–$0.040/minute by provider, telephony at $0.015/minute (varies by country), and phone numbers at $2/month, with optional per-minute add-ons (knowledge base, denoising, guardrails, PII removal, AI quality assurance) stacking on top. Retell's own advertised all-in range is $0.07–$0.31/minute. There is no mandatory subscription; 20 concurrent calls are included, then $8 per concurrent call per month. Retell pricing, accessed 2026-08-18.

Component pricing is transparent in its parts but requires you to assemble your own total — and the 4× spread in Retell's advertised range shows how much the configuration matters.

3. Hosting fee plus pass-through (example: Vapi)

Vapi's Build plan lists call hosting at $0.05/minute, with speech recognition, language models, and text-to-speech billed at provider cost on top — or shown as $0 on Vapi's bill when you bring your own API keys, in which case the providers bill you directly. Telephony is separate. Ten concurrent calls are included ($10/line/month beyond), and compliance is an add-on: HIPAA at $2,000/month, zero data retention at $1,000/month. Vapi pricing, accessed 2026-08-18.

The critical reading: $0.05/minute is not a call price — it is the hosting portion of one, and the complete cost cannot be computed from the public page alone.

4. Bundled tiers with platform fees (example: Bland AI)

Bland bundles the language model, speech recognition, and text-to-speech into one talk-minute rate, but varies the rate and its allowances by tier: Start $0.14/minute with no platform fee (100 calls/day, 10 concurrent), Build $0.12/minute plus $299/month (2,000 calls/day, 50 concurrent), Scale $0.11/minute plus $499/month (5,000 calls/day, 100 concurrent), Enterprise custom. Transfer minutes bill separately at $0.05/$0.04/$0.03 by tier (waived with bring-your-own telephony), and built-in telephony is pass-through. Bland pricing, accessed 2026-08-18.

The rate falls as the fixed fee rises, so the cheapest tier depends on your volume — and the daily-call and concurrency caps can force an upgrade the minutes alone would not justify.

5. Per-agent flat or per-call (examples: Goodcall, Smith.ai)

Receptionist-style products often drop per-minute billing entirely. Goodcall charges a flat $79, $129, or $249 per month per agent with, in its words, unlimited minutes and tokens — no fees for call counts, minutes, or tokens. The tiers instead cap logic flows, team members, and unique customers per month (100/250/500, then $0.50 per additional customer). Goodcall pricing, accessed 2026-08-19.

Smith.ai's AI Receptionist bills per call: a free plan with 25 calls/month ($3.00 per call after); Pro volume sub-tiers at $150, $270, and about $500/month covering 75, 150, and 300 calls respectively ($2.00/$1.80/$1.67 per call, overage $2.50/$2.30/$2.17); and Enterprise sub-tiers at $500 and $800/month covering 300 and 500 calls ($1.67/$1.60 per call, overage $2.17/$2.10), with 1,000+ calls custom-priced. Smith.ai AI Receptionist pricing, accessed 2026-08-19.

These models are predictable for steady, short-call workloads, but the unit — "call", "unique customer" — does the work the minute does elsewhere: a per-call price is cheap for two-minute calls and expensive for twenty-minute ones.

One month, five bills: 10,000 connected minutes

Take one explicit workload: 2,000 calls per month × 5 connected minutes = 10,000 connected minutes, evenly spread, peaking under 10 simultaneous calls, one included language, one number, no transfers, hold, or voicemail minutes. This is illustrative arithmetic, not a quote.

Sources: Retell, Vapi, Bland (accessed 2026-08-18); Goodcall, Smith.ai (accessed 2026-08-19); ThunderPhone published rates (2026-08-19).

Three readings. First, rankings do not survive workload changes: halve the call length and per-call vendors get relatively cheaper; double it and per-minute bills double while Goodcall's does not move. Second, identical-looking numbers mean different things — ThunderPhone's 5¢ Bolt rate is a finished call price, Vapi's $0.05 hosting fee is the platform's share of an unfinished one. Third, tiered platform fees can outweigh lower rates: Bland's Scale tier costs more than Start here, buying headroom rather than savings.

The cost drivers beyond the headline rate

Six things move the real bill in every model:

• Telephony and numbers. Carrier minutes and phone numbers usually sit outside the headline rate: Retell lists $0.015/minute and $2/number/month; Bland's built-in telephony is pass-through; Vapi prices telephony separately; ThunderPhone's is setup-dependent, with SIP and bring-your-own-number support. (Retell, Bland, accessed 2026-08-18.)
• Premium voices and languages. ThunderPhone's premium voice or language paths add up to 3¢/minute; Retell's text-to-speech line item runs $0.015–$0.040/minute by voice provider. Price the languages and voices you will actually use.
• Feature add-ons. Per-minute surcharges compound: Retell's knowledge base, denoising, guardrails, PII removal, and AI quality assurance each add to the rate; ThunderPhone publishes its acknowledgement, supervision, and long-prompt surcharges per minute.
• Concurrency. Simultaneous capacity is metered: Retell includes 20 concurrent calls ($8 each beyond), Vapi 10 ($10/line/month beyond), Bland caps by tier — a traffic spike can cost more in capacity than in minutes.
• Platform fees and compliance fees. Bland's Build and Scale tiers carry $299 and $499 monthly fees; Vapi lists HIPAA at $2,000/month and zero data retention at $1,000/month; Goodcall and Smith.ai are subscription-shaped by design. ThunderPhone and Retell pay-as-you-go list no mandatory subscription.
• Minimums, caps, and non-talk minutes. Smith.ai's Enterprise plan starts at a $500/month minimum; Bland caps calls per day; Goodcall caps unique customers; and non-conversation time — hold, voicemail, phone menus, transfers — can bill at its own rates (ThunderPhone publishes hold at a flat 2¢/minute and caps straight-to-voicemail billing at one minute; Bland bills transfer minutes by tier).

Total-cost-of-ownership checklist

Before committing, price your own traffic against each candidate:

1. Compute your real monthly connected minutes and call count — different models bill on different units.
2. Ask for the complete per-minute (or per-call) price of your exact configuration — language, voice, model, add-ons — not the headline rate.
3. Add telephony: carrier minutes, number fees, and whether bringing your own trunk changes the price.
4. Check every cap — calls/day, concurrency, unique customers, included calls — against your peak day, not your average.
5. Add fixed costs: platform fees, compliance fees, minimums, annual commitments.
6. Ask how hold, voicemail, phone menus, and transfers bill.
7. Budget testing minutes: simulated calls are often billable real calls (ThunderPhone shows the charge before a simulation run; Vapi's docs state test-suite calls bill like regular calls — Vapi Test Suites, accessed 2026-08-18).
8. Model a growth month: which bill line doubles when traffic doubles, and which does not.

FAQ

How much does an AI phone agent cost per minute?

Published complete per-minute prices run from about 2¢ to 31¢. ThunderPhone's all-in engine rates are 2¢/5¢/9¢; Retell advertises $0.07–$0.31/minute all-in; Bland's bundled tiers run $0.11–$0.14 per talk minute plus platform fees; Vapi's $0.05/minute is hosting only, before provider costs. (Retell, Vapi, Bland, accessed 2026-08-18.)

What does an AI phone agent cost per month for a small business?

In the 2,000-call, 10,000-minute scenario above, published arithmetic spans roughly $200 to $3,100+ per month — ThunderPhone $200–$900 by engine, Goodcall $79–$249 per agent flat (plus unique-customer overages), Vapi $500 hosting before pass-through, Retell $700–$3,100 advertised all-in, Bland $1,400 on its cheapest eligible tier, Smith.ai in custom-pricing territory at that call count. Rerun the arithmetic with your own numbers.

Are flat-rate "unlimited minutes" plans really unlimited?

The minutes can be, but another unit is usually metered. Goodcall states it charges no fees for calls, minutes, or tokens — its tiers instead cap unique customers per month (100/250/500, then $0.50 each), logic flows, and team members. The cap is the real meter. Goodcall pricing, accessed 2026-08-19.

Why is a per-call price hard to compare with a per-minute price?

Because call length converts one into the other. Smith.ai's published per-call rates of roughly $1.60–$2.00 equal 32–40¢/minute on a five-minute call but 80¢–$1.00/minute on a two-minute call. Compute your average call length before comparing the two. Smith.ai AI Receptionist pricing, accessed 2026-08-19.

What hidden costs should I expect beyond the advertised rate?

Telephony, premium voices or languages, add-on surcharges, concurrency, platform and compliance fees, and non-talk minutes. Published examples: Retell's per-minute add-ons and $8/concurrency/month beyond 20; Vapi's $2,000/month HIPAA add-on; Bland's $299/$499 platform fees and tiered transfer-minute rates. (Retell, Vapi, Bland, accessed 2026-08-18.)

Does compliance change the price?

On some platforms, yes — materially. Vapi lists HIPAA compliance at $2,000/month and zero data retention at $1,000/month as paid add-ons. ThunderPhone's public posture is GDPR and HIPAA compliant, and its published pricing lists no compliance add-on fee; customers remain responsible for configuring their workflows, data handling, recordings, integrations, and notices for their own obligations. Vapi pricing, accessed 2026-08-18.

Sources and freshness

All third-party rates on this page were checked against the linked vendor pages on the accessed dates shown beside each claim. ThunderPhone pricing and product details reflect its published pricing and documentation as of August 19, 2026. This guide is documentation-based — no vendor account, invoice, or live call was tested — and vendors change pricing frequently, so confirm against the live vendor pages before relying on them.