ThunderPhone pricing explained: 2¢, 5¢, and 9¢ engine rates, add-ons, and what a month actually costs

ThunderPhone uses prepaid, pay-as-you-go pricing: one per-minute engine rate covers speech recognition, the language model, and the voice. There is no subscription, seat fee, or platform fee, and the agent builder shows the all-in per-minute rate for the exact configuration before you put it into service.

That makes the base calculation direct, but it does not mean every possible configuration costs the base rate. Optional capabilities, long prompts, some voice-language combinations, demo-number usage, and telephony can change the total. Billing History itemizes every call, so the rate shown in the builder can be reconciled with actual usage.

The three ThunderPhone engine rates

The engines are priced by the kind of workload they are designed to handle. Speech recognition, language-model processing, and voice output are included in each base rate rather than appearing as three separate component charges.

The right starting point is therefore the engine that fits the job, not simply the lowest number. Once an engine and options are selected, the builder's displayed rate is the useful number for budgeting that specific agent.

What's included and what bills on top

ThunderPhone does not charge a separate platform fee for the core platform capabilities used to build and operate an agent. The distinction to watch is between capabilities available as part of the platform and options that add a published per-minute surcharge.

ThunderPhone has 123 voices across 47 languages. Fifteen languages are included, while the other 32 are premium at up to 3¢/minute. Agents can switch languages during a call when additional languages are configured. Where segment-level language billing is available, the language actually spoken in each segment determines the charge.

The practical rule is to configure the agent first and read the displayed all-in rate. In particular, a language label and a premium-voice label do not automatically create two surcharges. The effective speech surcharge comes from the actual pairing.

Metering mechanics

Calls are metered in quarter-minute increments, with a minimum of 0.25 billable minutes. The rate and options active when the call occurs determine its charge. Editing an agent later does not rewrite Billing History.

Not every call state uses the same rate:

• Hold time bills at a flat 2¢/minute on Spark, Bolt, and Storm, regardless of how long the queue runs.
• Voicemail and phone-menu time bills at the selected engine rate. A call that goes straight to voicemail is capped at one billable minute.
• Voicemail handling can let the agent prompt decide, hang up, or leave a configured message. Those choices control behavior; the voicemail billing rule remains the same.
• Language switching can follow the language actually spoken by segment where that billing information is available, rather than treating every segment as though it used every configured language.

These details matter when connected conversation minutes are only part of the traffic. A forecast should separate ordinary conversation, hold, voicemail, and phone-menu time rather than multiplying all call duration by one assumed rate.

Worked examples: what a month costs

Every calculation in this section is illustrative, not a quote. It uses published rates and excludes setup-dependent telephony. A selected language or premium voice path can add up to 3¢/minute unless the example explicitly includes it.

Small-business month: 1,200 connected minutes

Assume 300 calls, each with 4 connected minutes. Illustrative arithmetic, not a quote:

• Usage: 300 calls × 4 minutes = 1,200 minutes
• Spark: 1,200 × $0.02 = $24.00
• Bolt: 1,200 × $0.05 = $60.00
• Storm: 1,200 × $0.09 = $108.00

There is no subscription to add. The resulting illustrative engine spend is $24 on Spark, $60 on Bolt, or $108 on Storm before optional additions and telephony.

Standard month: 10,000 connected minutes

Now assume 2,000 calls, each with 5 connected minutes. Illustrative arithmetic, not a quote:

• Usage: 2,000 calls × 5 minutes = 10,000 minutes
• Spark: 10,000 × $0.02 = $200.00
• Bolt: 10,000 × $0.05 = $500.00
• Storm: 10,000 × $0.09 = $900.00

Again, these are base-engine totals. Hold, optional additions, and setup-dependent telephony must be modeled separately.

Configured Storm example: acknowledgements and a premium language

Suppose a Storm agent uses verbal acknowledgements and a voice-language combination carrying the maximum 3¢/minute speech surcharge. Illustrative arithmetic, not a quote:

• Configured rate: $0.09 + $0.02 + $0.03 = $0.14/minute
• Monthly usage: 2,000 calls × 5 minutes = 10,000 minutes
• Engine and selected additions: 10,000 × $0.14 = $1,400.00

The example shows why the builder's configured rate matters. The 14¢ rate is the sum of the selected published items, not a separate plan or universal Storm price.

Prompt-size example: a 15,001-token Storm prompt

The first 5,000 estimated prompt tokens are included. The remaining 10,001 tokens occupy one full 10,000-token block and part of a second, so two additional blocks bill. Illustrative arithmetic, not a quote:

• Tokens above the included amount: 15,001 − 5,000 = 10,001 tokens
• Storm prompt addition: 2 blocks × $0.02 = $0.04/minute
• Storm rate with this prompt addition and no other add-ons: $0.09 + $0.04 = $0.13/minute

The partial second block is why a prompt just one token over 15,000 produces two blocks rather than one. Prompt estimates can therefore matter near a block boundary.

Getting started, balance controls, and phone numbers

Signing up and building agents is free. To make real calls, add a payment method and at least $3.00 of prepaid balance. Funds do not expire.

Auto top-up is optional. It can be configured with a balance threshold, a refill amount, and a monthly spending limit; alerts fire at 80% and 100% of that limit. Those controls help prevent an active agent from silently outrunning a planned monthly budget while still letting prepaid usage continue when intended.

For testing, a US demo number costs $1/month and adds 1¢/minute when it carries a call. Demo numbers are inbound-only, limited to 10 per organization, and intended for testing rather than production. Imported VoIP/SIP numbers have no ThunderPhone import or monthly charge, but the VoIP provider continues charging for its number and telephony usage. That provider cost is why telephony is setup-dependent rather than included in the engine rate.

ThunderPhone is GDPR and HIPAA compliant, and customers remain responsible for configuring their workflows, data handling, recordings, and notices for their own obligations. No compliance add-on fee appears in published ThunderPhone pricing.

How testing and simulations bill

ThunderPhone supports browser microphone tests, AI-caller simulations, scenario suites, regression validation sets, and live A/B experiments. AI-caller simulations are billable real calls, and the launch dialog shows the charge before a run.

In a simulation, the simulated caller bills at the Spark rate while the agent under test bills at its actual configured rate. This makes the preview important: a test involves both sides of a real simulated conversation, not merely a free editor action. Testing workflows should be included in usage planning when scenario suites or regression sets run regularly.

How this model compares with other pricing structures

Some voice-AI products use subscriptions with included minutes, some publish a hosting charge while passing component costs through, and others assemble a total from separate line items. The voice AI pricing comparison normalizes those structures, while the AI phone-agent cost guide explains the major pricing models and the cheapest AI voice-agent guide examines defined cost scenarios.

Compare complete configured costs, not isolated headline numbers. Call length, language, and configuration can reorder results, so rerun the arithmetic for the workload you expect.

FAQ

How much does ThunderPhone cost per minute?

The base engine rates are 2¢/minute for Spark, 5¢/minute for Bolt, and 9¢/minute for Storm. Speech recognition, the language model, and voice are included. Optional additions, long prompts, hold-time rules, and setup-dependent telephony can change the total for a particular call.

Is there a subscription or platform fee?

No. ThunderPhone has no subscription, seat fee, or platform fee. It uses prepaid, pay-as-you-go billing, and Billing History itemizes every call at the configuration active when that call occurred.

What does the 2¢ rate include?

Spark's 2¢/minute rate includes speech recognition, the language model, and voice. It is intended for cost-efficient, straightforward workflows. A selected language or premium voice path, supervision, demo-number usage, live caller transcripts, or a long prompt can add to the base rate when configured.

What costs extra?

Published additions include Storm acknowledgements, Storm extra intelligence, supervision, selected language or premium voice paths, demo-number usage, live caller transcripts, and prompt blocks beyond the included size. Telephony depends on setup, and hold bills at a separate flat rate. The builder displays the all-in per-minute rate for the exact agent configuration.

Is there a free trial?

ThunderPhone is free to sign up for and free to use while building agents; published pricing does not define a separate trial allowance. Real calls require a payment method and at least $3.00 of prepaid balance. Funds do not expire, and simulations are billable real calls with the charge shown before launch.

How does billing work for hold time and voicemail?

Hold time bills at a flat 2¢/minute on every engine; voicemail and phone-menu minutes bill at the engine rate. Straight-to-voicemail calls are capped at one billable minute. For voicemail behavior, the agent can let its prompt decide, hang up, or leave a configured message.

Sources and freshness

ThunderPhone pricing and product figures on this page reflect the published ThunderPhone pricing page and ThunderPhone pricing documentation, accessed 2026-09-01. This guide explains published pricing rather than reproducing an invoice or providing a quote. Confirm the builder's displayed all-in rate for the exact engine, language, voice, prompt size, and options you plan to use.