Voice AI pricing compared (2026): ThunderPhone vs. Retell vs. Vapi vs. Bland per-minute costs

Voice AI platforms quote per-minute prices, but the prices do not describe the same thing. One vendor's headline rate is a complete call price; another's is an infrastructure line item that grows once a language model and voice are attached; a third's is a hosting fee that by design excludes the speech, model, and voice components; a fourth's is a bundled rate gated behind daily-call and concurrency limits. A buyer who compares the four headline numbers directly will rank the platforms incorrectly.

This guide explains each pricing model in its own terms, then normalizes all four against one explicit workload. ThunderPhone figures reflect its published rates verified September 10, 2026. Competitor figures were re-checked against first-party pricing pages that day, except where an earlier access date is retained beside the claim. This is a neutral buyer's guide, and none of the arithmetic below is a quote.

Four pricing models, briefly

ThunderPhone: all-in engine rates

ThunderPhone prices the complete engine — speech recognition, language model, and voice — as one per-minute rate: Spark at 2¢/minute, Bolt at 5¢/minute, and Storm at 9¢/minute, pay as you go with no subscription or seat fee. Premium voice or language paths add up to 3¢/minute (non-stacking), and optional add-ons carry published surcharges: verbal acknowledgements +2¢ (Storm only) and call supervision +4¢ (Spark and Storm), long prompts +1–2¢/minute per additional 10K tokens after the first 5K, and ThunderPhone numbers +1¢/minute plus $1/month. Hold time bills a flat 2¢/minute on every engine; 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 for the exact configuration, and telephony charges depend on setup.

Retell AI: component pricing with an advertised range

Retell prices components separately: voice infrastructure at $0.055/minute, language models from $0.003 to $0.16/minute depending on the model, text-to-speech at $0.015–$0.040/minute by provider, US telephony at $0.015/minute in its Twilio example. The advertised range is $0.07–$0.31/minute; its headline floor excludes the LLM and telephony. There is no mandatory subscription on pay as you go; 20 concurrent calls are included and added concurrency is $8 per concurrent call per month. Retell pricing, accessed 2026-09-10.

Retell also lists phone numbers at $2/month. Optional add-ons stack on top: knowledge base, denoising, and guardrails at $0.005/minute each, PII removal at $0.01/minute, and AI quality assurance at $0.10/minute after the first 100 free minutes. Retell pricing, accessed 2026-08-18 for number rental and these add-on rates.

Vapi: hosting fee plus provider pass-through

Vapi's Build plan lists call hosting at $0.05/minute. Speech recognition, language models, and text-to-speech bill at provider cost on top — or show as $0 on the Vapi bill when customers bring their own API keys, in which case the providers bill them directly. Telephony is separate. Ten concurrent calls are included, with extra lines at $10 each per month; the Scale plan uses a fixed platform fee with committed volume and custom terms. Compliance is also an add-on: HIPAA at $2,000/month and zero data retention at $1,000/month. The critical point: $0.05/minute is not a call price — it is the hosting portion of one. Vapi pricing, accessed 2026-09-10.

Bland AI: tiered bundles with platform fees and limits

Bland prices an all-inclusive talk minute — the rate covers the language model, speech recognition, and text-to-speech with no token charges or model-provider pass-through — but the rate and its operational allowances vary by tier. Start is $0.14/minute with no platform fee, limited to 100 calls/day and 10 concurrent calls; Build is $0.12/minute plus $299/month with 2,000 calls/day and 50 concurrent; Enterprise is custom. Scale is no longer published. Transfer minutes bill separately at $0.05/minute on Start and $0.04/minute on Build (the page stated on 2026-08-18 that customers who bring their own telephony do not pay transfer fees; confirm this still applies); built-in Twilio telephony is at pass-through cost. Bland pricing, accessed 2026-09-10.

Other published models: Plivo fits engine-style pricing at $0.03/minute before telephony, although LLM inclusion is unconfirmed. Plivo pricing, accessed 2026-09-10. Telnyx fits component pricing: its $0.05/minute engine excludes LLM and telephony. Telnyx pricing, accessed 2026-09-10. Deepgram Standard PAYG bundles the engine at $0.056/minute through September 12, 2026, then $0.075/minute from September 13; carrier costs remain separate. Deepgram pricing, accessed 2026-09-10. Cartesia combines a $0.06/minute agent rate with component charges and plan limits; its UI-agent LLM offer is temporary. Cartesia pricing, accessed 2026-09-10. Twilio ConversationRelay resembles hosting plus pass-through: $0.07/minute before voice and the customer’s LLM; STT/TTS inclusion is unconfirmed. ConversationRelay pricing, accessed 2026-09-10. ElevenLabs uses tiered allowances with $0.08/minute standard overage, plus LLM and telephony. ElevenLabs Agents pricing, accessed 2026-09-10. The Voice AI Price Index records the comparable floors and remaining uncertainties.

At a glance

One workload, four bills

Use one explicit workload: 2,000 calls per month × 5 connected minutes = 10,000 connected minutes, spread evenly, peaking at no more than 10 simultaneous calls. Assume one primary included language, ordinary connected talk time, one phone number, no transfer minutes, and no voicemail, hold, or screening minutes. This is illustrative planning arithmetic, not a quote.

ThunderPhone — the engine rate is the complete base-engine usage price:

• Spark: 10,000 × $0.02 = $200/month
• Bolt: 10,000 × $0.05 = $500/month
• Storm: 10,000 × $0.09 = $900/month
• A premium voice or language path at the maximum 3¢/minute would add up to $300/month; telephony is setup-dependent and not added here. Subscription: $0.

Retell AI — the advertised range applied to the workload:

• Low end: 10,000 × $0.07 = $700/month
• High end: 10,000 × $0.31 = $3,100/month

The headline floor excludes the LLM and telephony. A configured floor computes as 10,000 × ($0.055 infrastructure + $0.015 voice + $0.003 LLM + $0.015 telephony) = $880/month, before number rental and add-ons. ThunderPhone's Storm at $900 falls inside Retell's advertised range; Spark ($200) and Bolt ($500) are below this configured Retell floor. ThunderPhone telephony remains extra. Retell pricing, accessed 2026-09-10.

Vapi — the public arithmetic covers hosting only:

• Call hosting: 10,000 × $0.05 = $500/month — before speech, model, voice, and telephony costs, which pass through at provider rates or bill directly under your own API keys.

The complete total is $500 + provider costs + telephony and cannot be computed from the public page. It is worth pausing on the coincidence: ThunderPhone's Bolt all-in rate (5¢) equals Vapi's hosting-only fee ($0.05) — one is a complete engine rate, the other is the platform's share of an incomplete total. The 10-call peak stays within Vapi's 10 included concurrent calls, so no extra lines are needed. Vapi pricing, accessed 2026-09-10.

Bland AI — the scenario stays within 100 calls/day and peaks at 10 concurrent, so it fits Start's published limits if traffic actually distributes that way:

• Start: 10,000 × $0.14 + $0 = $1,400/month
• Build: 10,000 × $0.12 + $299 = $1,499/month

Build costs more at this volume because its platform fee outweighs its lower per-minute rate; it buys more call capacity. Built-in telephony is pass-through and is not added here. Bland pricing, accessed 2026-09-10.

Summary for this scenario: ThunderPhone $200–$900 by engine; Retell $880 at the configured floor before number rental and add-ons; Vapi $500 hosting before pass-through; Bland $1,400 on its cheapest eligible tier. Rerun the arithmetic with your own call length, language, and configuration before drawing a conclusion — average call length alone can reorder these results.

What a "minute" includes

The same word covers different bundles:

• ThunderPhone: the engine rate includes speech recognition, the language model, and the voice. Premium voice or language paths (up to +3¢/minute) and published add-on surcharges are the exceptions, shown in the builder as a current all-in rate. Hold bills a flat 2¢/minute; voicemail and phone-menu minutes bill at the engine rate, straight-to-voicemail capped at one minute.
• Retell: the $0.055/minute infrastructure rate includes none of the model or voice cost; each is its own line item, as are telephony and the optional per-minute add-ons. The $0.07–$0.31 advertised range does not establish a complete call price; its floor excludes LLM and telephony. Retell pricing, accessed 2026-09-10.
• Vapi: the $0.05/minute hosting fee includes none of the speech, model, or voice cost by design — those pass through at cost, or move onto your own provider bills with your own API keys. Vapi pricing, accessed 2026-09-10.
• Bland: the talk-minute rate includes the language model, speech recognition, and text-to-speech, with no token charges. Transfer minutes bill separately by tier, and telephony is pass-through. Bland pricing, accessed 2026-09-10.

Hidden-cost checklist for buyers

Before signing anything, price these against your actual traffic:

1. Non-talk minutes. Hold, voicemail, phone-menu navigation, and transfer time can bill differently from conversation. ThunderPhone publishes hold at a flat 2¢/minute and caps straight-to-voicemail billing at one minute; Bland bills transfer minutes at separate tiered rates. Ask every vendor how each call state bills.
2. Telephony. Carrier minutes and phone numbers are usually outside the headline rate — Retell lists $0.015/minute in its US Twilio example and $2/number/month (Retell pricing, accessed 2026-08-18 for number rental), Bland's built-in telephony is pass-through, Vapi prices it separately, and ThunderPhone's is setup-dependent. Retell pricing and Bland pricing, accessed 2026-09-10.
3. Concurrency and volume caps. Retell includes 20 concurrent calls ($8 each beyond), Vapi includes 10 ($10/line beyond), and Bland caps calls/day and concurrency by tier — a spike can force a tier upgrade that costs more than the minutes. Retell pricing, Vapi pricing, and Bland pricing, accessed 2026-09-10.
4. Platform and compliance fees. Bland's Build tier carries a $299 monthly fee; Start has no platform fee; Vapi lists HIPAA at $2,000/month and zero data retention at $1,000/month. ThunderPhone and Retell list no mandatory subscription on pay as you go. Vapi pricing and Bland pricing, accessed 2026-09-10.
5. Quality and 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's verbal acknowledgements, call supervision, and long-prompt surcharges are published per minute. Price the configuration you will actually run.
6. Testing minutes. ThunderPhone simulations are billable real calls, and Vapi's docs state test-suite calls bill like regular calls. Budget for a regression suite, not just production traffic. Vapi Test Suites, accessed 2026-08-18.
7. Who bills you. Bringing your own API keys to Vapi moves provider costs off Vapi's invoice onto the providers' — it changes who bills, not necessarily what the workload costs. Vapi pricing, accessed 2026-09-10.

Use it from a coding agent

As of September 2026 (checked September 16). These are documented development surfaces, not a test of call quality. A hosted account MCP lets a coding agent operate the platform; a docs MCP helps it read documentation. Unknown means the inspected sources did not establish the capability, not that the vendor lacks it.

ThunderPhone's documented tools cover agent creation, configuration, test scenarios, test runs, calls, and imports. Use tools/list to discover the tools enabled for your key. Counts are not a measure of workflow quality. See the agent-readiness index for docs MCP, directory status, CLI support, and dated sources for each cell.

Try ThunderPhone:

npx skills add thunderphone/skills
MCP: https://api.thunderphone.com/v1/mcp
Setup and examples: /developers

FAQ

Which platform is cheapest for 10,000 connected minutes per month?

In this defined included-language scenario, ThunderPhone Spark has the lowest published complete engine arithmetic. Its engines calculate to $200–$900; Retell's configured floor is $880 before number rental and add-ons; Vapi's knowable public figure is $500 for hosting alone, before provider and telephony costs; Bland's cheapest eligible tier is $1,400. Different call lengths, languages, and add-ons can change the ordering, so rerun the arithmetic with your own configuration. Retell pricing, Vapi pricing, and Bland pricing, accessed 2026-09-10. A broader ranking across more platforms — including Telnyx, per-call, and per-customer products — is in the cheapest AI voice agent guide.

Why can't I just compare the headline per-minute rates?

Because they price different bundles. ThunderPhone's 5¢ Bolt rate is a complete engine price, while Vapi's $0.05 hosting fee excludes speech, model, voice, and telephony; Retell's $0.055 infrastructure rate similarly excludes model and voice costs, and its $0.07 headline floor still excludes LLM and telephony. Normalize to a complete configured minute before comparing. Retell pricing and Vapi pricing, accessed 2026-09-10.

Which platforms charge a monthly platform fee?

Bland's Build tier does ($299/month); ThunderPhone, Retell pay-as-you-go, and Vapi's Build plan list no mandatory subscription. Vapi's Scale plan uses a fixed platform fee with committed volume, and compliance add-ons (HIPAA $2,000/month, zero data retention $1,000/month) function like recurring fees where required. Bland pricing and Vapi pricing, accessed 2026-09-10.

How do voicemail, hold, and transfer minutes bill?

Differently on each platform, and the details matter at volume. ThunderPhone bills hold at a flat 2¢/minute and voicemail and phone-menu minutes at the engine rate, with straight-to-voicemail calls capped at one minute. Bland bills transfer minutes at $0.05/minute on Start and $0.04/minute on Build (the page stated on 2026-08-18 that customers who bring their own telephony do not pay transfer fees; confirm this still applies). The reviewed Retell and Vapi public pages do not itemize these states, so ask directly. Bland pricing, accessed 2026-09-10.

Does bringing my own API keys make Vapi cheaper?

It makes Vapi's invoice smaller, not necessarily the workload cheaper. Vapi lists speech, model, and voice as $0 on its bill when customers bring their own keys, but usage under those keys is billed by the providers at their rates. The comparison that matters is total spend across all invoices for the same configured call. Vapi pricing, accessed 2026-09-10.

What should I check before trusting any of these numbers?

The vendor's live pricing page, against your own traffic profile. Voice AI pricing changes frequently; every third-party figure here is a point-in-time reading with its accessed date attached, the scenario assumes an even traffic distribution that real workloads may not match, and telephony is excluded except where a configured example explicitly adds it. Reproduce the arithmetic from current sources before deciding.

Sources and freshness

First-party pricing figures were re-checked on September 10, 2026. ThunderPhone pricing and product details were verified that day. Bland now publishes Start, Build, and custom Enterprise; the former Scale tier is removed. Retell number rental and optional add-on rates carry their 2026-08-18 verification. Bland’s bring-your-own-telephony transfer-fee waiver and Vapi’s Test Suites documentation also retain their 2026-08-18 access dates.

• ThunderPhone pricing, accessed 2026-09-10.
• Retell pricing, accessed 2026-09-10; number rental and optional add-on rates accessed 2026-08-18.
• Vapi pricing, accessed 2026-09-10.
• Vapi Test Suites, accessed 2026-08-18.
• Bland pricing, accessed 2026-09-10; bring-your-own-telephony transfer-fee waiver accessed 2026-08-18.
• Plivo pricing, accessed 2026-09-10.
• Telnyx pricing, accessed 2026-09-10.
• Deepgram pricing, accessed 2026-09-10.
• Cartesia pricing, accessed 2026-09-10.
• Twilio ConversationRelay pricing, accessed 2026-09-10.
• ElevenLabs Agents pricing, accessed 2026-09-10.

This comparison is documentation-based; no vendor account, invoice, or live call was tested. Vendors change pricing often, so confirm against the live vendor pages before relying on any figure here.

Video: the same comparison, with every pricing page on screen

The embedded video above walks through each vendor's public pricing page (read on 10 September 2026) and adds up what one live minute costs. Full transcript:

0:00 The question and the method

Voice AI pricing pages all look different, so let's put the platforms side by side and work out what one minute of a live phone call actually costs on each of them. The method is simple. For every platform we open its public pricing page, as of September tenth, twenty twenty-six, and add up the platform fee plus the three models every call needs: speech to text, the language model, and text to speech. Telephony, monthly fees, and compliance add-ons get noted separately. We are comparing price only. Whether the calls actually work is a separate question, and a separate video.

0:35 Vapi: 5¢ platform fee, models billed on top

First, Vapi. The Build plan is usage based, and the line that matters is here: Vapi hosting cost, five cents per minute, and it explicitly excludes model provider costs. Speech to text, the language model, and text to speech are billed at cost on top, or you bring your own API keys and pay those vendors directly. Concurrency is ten calls included, then ten dollars per line per month. HIPAA is a two thousand dollar a month add-on, and zero data retention is a thousand a month. So the five cents is the floor, not the price of a minute. A typical stack with models and telephony lands around seven cents and up.

1:11 Retell: 7¢ to 31¢, their own example is 11.5¢

Retell publishes a range: seven to thirty-one cents per minute for voice agents, pay as you go, with ten dollars of free credit. Their own cost estimator shows where a minute lands. With the default choices, the language model is four and a half cents, Retell's voice infrastructure is five and a half cents, text to speech is a cent and a half, and telephony is zero only because it assumes you bring your own. That adds up to eleven and a half cents per minute before the phone line.

1:38 Bland: 14¢, or 12¢ plus $299 a month

Bland goes the other way and bundles the models. The Start plan is fourteen cents per minute with no platform fee, but it is capped at one hundred calls a day. The Build plan drops to twelve cents per minute and adds a two hundred ninety-nine dollar monthly platform fee. Transfer time bills separately at four or five cents.

1:56 Telnyx: 5¢ engine, LLM and telephony on top

Telnyx says five cents per minute, no fine print, and it is the most complete of the itemized offers: orchestration, speech to text, and text to speech are all inside the five cents. The language model is an add-on, and the headline figure, about four tenths of a cent per minute, is for Kimi, an open-source model on Telnyx's own GPUs, not a frontier model. Frontier models from OpenAI or Anthropic are billed at those vendors' rates, so a production agent on a frontier model costs more. Telephony is carrier cost, from about a third of a cent. Their own estimator, with the open-source model, lands at about five point six cents per minute including telephony. Volume discounts start at a five hundred dollar monthly commitment.

2:41 Synthflow: enterprise only, from $30,000 a year

Synthflow no longer publishes a per-minute rate at all. The pricing page has one plan, enterprise, with contracts starting at thirty thousand dollars a year and final pricing scoped to volume. Fine for a large contact center, out of reach for a self-serve team.

2:57 ThunderPhone: 2¢, 5¢, 9¢ all in, no platform fee

Now ThunderPhone. Three tiers, each one flat rate with the models included: Spark at two cents per minute, Bolt at five, Storm at nine. Models included means all of them: ThunderPhone runs several models that check each other rather than one transcription model, one language model, and one voice, and that whole stack is inside the rate. There is no platform fee, no monthly fee, and no per-line concurrency charge. Telephony is either your own SIP trunk, connected for free, or a US phone number from ThunderPhone at a dollar a month plus one cent per minute. Hold time bills at two cents a minute on every tier, and the optional add-ons are listed with their prices on the page. And for high-volume enterprise use, ThunderPhone goes lower still: one cent a minute or less under an enterprise agreement, which is what makes high-volume, low-margin outbound work.

3:40 Scoreboard: cost per live minute

Here is the scoreboard, platform plus models, per live minute, telephony noted separately. ThunderPhone Spark, two cents. Bolt, five. Storm, nine, with all the models included. Telnyx, about five point four cents once you add the open-source language model. Vapi, five cents plus whatever the models cost. Retell, seven to thirty-one cents, eleven and a half in their own example. Bland, fourteen cents, or twelve plus the monthly fee. Synthflow, enterprise contracts only. Even ThunderPhone's most capable tier, Storm, is at or below most competitors' entry rates, and the cheapest tier is less than half of the next cheapest platform.

4:18 What a 10,000-minute month costs

In dollars, for a team doing ten thousand live minutes a month: two hundred dollars on ThunderPhone Spark, five hundred on Bolt, nine hundred on Storm. Around five hundred forty on Telnyx. Five hundred on Vapi before the model bills. Eleven hundred fifty on Retell using their example. Fourteen hundred on Bland Start, if the daily cap even allows it, or fourteen ninety-nine on Bland Build.

4:41 Why ThunderPhone can be the cheapest

Why the gap? ThunderPhone does not charge a platform fee on top of model costs, and it tunes its stack for performance per dollar. All of these numbers come from the vendors' own public pricing pages, linked in the description with the date they were read. Prices change, so check the links. As of today, ThunderPhone is the cheapest voice AI platform, from two cents a minute, all in.