Free data · updated monthly

AI voice agent pricing tracker

Published voice agent platform rates start at $0.038 a minute (Amazon Connect Customer, telephony extra). Bland’s $0.14 Start rate bundles the model and voice, and Retell publishes a $0.07 to $0.31 range for voice agents. Most low headline rates cover one layer: Vapi’s $0.05 and Retell’s $0.055 add the model, voice and telephony on top.

Last checked Next check November 2026

Every price is a list price in US dollars, read from the vendor’s own page, with the source and date on each row. Apptycoons builds custom voice agents. No affiliate links, ratings or rankings. Data: CSV · JSON.

Voice agent platform prices, checked October 2026

The headline rates below measure different things. Some cover only orchestration, some bundle the model and voice, and almost none include the phone call. Read the “included” and “billed separately” columns before comparing two numbers. Vendors are listed in no particular order; this is not a ranking.

11 platforms · list prices in USD · checked October 19, 2026
PlatformPricing modelHeadline rateWhat’s includedBilled separatelyFree tierSource
VapiPer-minute hosting fee, optional monthly package$0.05/min hostingCore $29/month; Pro 10% of hosting fee, $999/month minimum; Premier custom
  • Orchestration (Vapi hosting)
  • Vapi telephony and SIP, WebSockets and Daily WebRTC: free
  • Speech-to-text, model and voice: passed through at provider cost, no markup
  • Twilio, Vonage or Telnyx minutes: charged by the provider
$5 credit, 1 Vapi number, 4 concurrent callsvapi.ai
Retell AIPer minute, priced by component$0.055/min voice infrastructure ($0.07–$0.31/min for voice agents, Retell’s published range)Pay as you go; enterprise at custom pricing
  • Voice infrastructure
  • First 20 concurrent calls
  • First 10 knowledge bases stored ($80/month value)
  • Voice: $0.015/min platform voices, up to $0.10/min for ElevenLabs v3 and v4
  • Model: e.g. Claude 4.5 Haiku $0.025/min, Claude 5 Sonnet $0.064/min
  • Telephony: $0.015/min on US local numbers (Twilio), $0.06/min toll-free; Retell numbers $2/month
  • Knowledge base use: +$0.005/min
$10 creditretellai.com
Bland AIPer talk minute, optional monthly platform feeStart $0.14/min; Build $0.12/min + $299/monthStart $0 platform fee; Build $299/month; Enterprise contracted to volumeModel, speech-to-text and voice (no token charges)
  • Telephony: your carrier, or Bland’s Twilio at pass-through cost
  • Transfers: $0.05/min (Start), $0.04/min (Build); none on your own telephony
Start plan: $0 platform fee, 2 credits and an inbound numberbland.ai
ElevenLabsElevenAgentsMonthly plan with included call minutes, then per minute$0.08/min additional call minutes ($0.16/min burst, when enabled)Starter $6, Creator $22, Pro $99 (1,238 min), Scale $299, Business $990 a month; Enterprise custom
  • Text-to-speech, speech-to-text, knowledge bases and RAG
  • Included minutes: 15 (Free) to 12,375 (Business)
  • Model (LLM): billed separately by model
  • Carrier charges: your telephony provider bills you directly
15 call minutes a monthelevenlabs.io
DeepgramVoice Agent APIPer minute of WebSocket connection$0.075/min Standard, pay as you go ($0.068/min Growth)Pay as you go: Advanced $0.163/min ($0.146 Growth); Standard with your own voice (BYO TTS) $0.065/min, Advanced BYO TTS $0.122/min; your own model (BYO LLM) $0.059/min; your own model and voice $0.050/min. Growth from $4K a yearSpeech-to-text, model and voice in one API (Standard tier)Telephony: not covered on the pricing page$200 credit for new accountsdeepgram.com
TwilioConversationRelayPer minute, plus Twilio voice minutes$0.07/minThe ConversationRelay layer between a Twilio call and your AI app; speech and voice are not itemized
  • Voice (call) minutes: billed separately, e.g. $0.0085/min US inbound local
  • Your language model
Free trial, no card requiredtwilio.com
CartesiaManaged AgentsMonthly plan with prepaid agent usage, then per minute$0.06/min call durationPro $5, Startup $49, Scale $299 a month, each as prepaid agent usage; Enterprise custom
  • Agent call minutes
  • Model use for agents created in the UI: free for a limited time
Telephony: $0.014/min on a Cartesia-provided number$1 of agent usage a monthcartesia.ai
Amazon Web ServicesAmazon Connect CustomerPer voice minute; AWS lists no token charges$0.038 per voice minuteAmazon Connect Customer voice minutes (AWS lists no token charges)
  • Telephony: standard telephony rates apply
  • Additional capabilities: AWS points to a separate pricing-details page
No free allowance stated on the pageaws.amazon.com
Google CloudDialogflow CX (Conversational Agents)Per billed second of audio$0.001/s with Flows, $0.002/s with PlaybooksVoice agent turns: listening time plus spoken-response time, rounded up per turnTelephony: not on the pricing page$600 (Flows) or $1,000 (Playbooks) credit, expires after 12 monthscloud.google.com
SynthflowSynthflow EnterpriseAnnual enterprise contractFrom $30,000 a yearTelephony or SIP, integrations, onboarding and launch support, scoped in the contractNot itemized on the pageNone listedsynthflow.ai
PolyAIPer minute (figure not published)No public price (sales quote)Per PolyAI: performance improvements, maintenance and 24/7 supportNot itemized on the pageNone listedpoly.ai

“Not stated on the pricing page” means we did not find it there, not that the vendor lacks it. Notes on individual rows (billing by connection time, promotional prices, contradictions on a vendor’s page) are in the JSON file.

Concurrency, HIPAA and other fixed fees

Concurrent calls and HIPAA terms are fixed costs that a per-minute rate doesn’t show, and vendors publish them in different places. Here is what each platform page says.

PlatformConcurrent callsHIPAA / BAASource
Vapi4, 10 or 30 by package; extra lines $10/month eachHIPAA add-on with BAA: $2,000/monthvapi.ai
Retell AI20 included; $8/month each above thatHIPAA/BAA on pay-as-you-go and enterprise in Retell’s plan table; custom BAA on enterpriseretellai.com
Bland AI10 (Start), 50 (Build), custom (Enterprise)BAA listed on Enterprise (custom price)bland.ai
ElevenLabsElevenAgents4 to 40 by plan; burst (when enabled) up to 3x at $0.16/minBAAs on Enterprise (custom price)elevenlabs.io
DeepgramVoice Agent APIUp to 45 connections (pay as you go), 60 (Growth)BAA for Enterprise customersdeepgram.com
TwilioConversationRelayNot stated on the pricing pageBAA needs Security or Enterprise Edition (price not published)twilio.com
CartesiaManaged Agents8, 12, 20 or 60 by planBAAs on Enterprisecartesia.ai
Amazon Web ServicesAmazon Connect CustomerNot stated on the pricing pageNot stated on the pricing pageaws.amazon.com
Google CloudDialogflow CX (Conversational Agents)Not stated on the pricing pageNot stated on the pricing pagecloud.google.com
SynthflowSynthflow EnterpriseScoped in the contractNot priced on the pagesynthflow.ai
PolyAINot stated on the pricing pageNot stated on the pricing pagepoly.ai

Model, speech and telephony prices

These are the published inputs that platform rates sit on, and what a custom stack pays directly. Each is shown in the vendor’s own unit. We don’t convert token or character prices to a per-minute figure unless the vendor publishes one.

ProductWhat it isPublished priceSource
TwilioProgrammable Voice (US)Telephony, per minute + number per monthInbound local $0.0085/min + $1.15/month number; toll-free $0.0220/min + $2.15/month; outbound $0.0140/min; Media Streams $0.0044/mintwilio.com
DeepgramNova-3 streamingSpeech-to-text, per minute$0.0077/min regular price ($0.0048/min current promotional price), pay as you godeepgram.com
DeepgramAura-2Text-to-speech, per 1K characters$0.030 per 1K characters, pay as you godeepgram.com
ElevenLabsText to Speech APIText-to-speech, per 1K charactersv4 Turbo and v3 Conversational $0.04 per 1K characters; v4, v3 and Multilingual v2 $0.08elevenlabs.io
CartesiaSonic text to speechText-to-speech, credits (1 credit = 1 character)Startup $49/month for 1.25M credits; overage $65, $45 or $38 per 1M credits (Pro, Startup, Scale)cartesia.ai
OpenAIRealtime API and GPT-LiveSpeech-to-speech model, per 1M tokens (gpt-live-1: per minute)gpt-realtime-2.1: audio $32 in, $64 out per 1M tokens; gpt-realtime-2.1-mini: $10 in, $20 out; gpt-live-1 voice sessions: $0.05/min, billed per second (backend model and tools extra)developers.openai.com
GoogleGemini Live APISpeech-to-speech model, per 1M tokens or per minuteGemini 3.8 Live: audio in $3.00 per 1M tokens or $0.005/min; audio out $12.00 or $0.018/minai.google.dev
AnthropicClaude APILanguage model, per 1M tokensSonnet 5.5 $2 in, $10 out; Haiku 5.5 $0.10 in, $0.50 out (prompts up to 100K tokens)claude.com

What platform prices leave out

A headline rate is one line of a five-line bill: telephony, speech-to-text, the language model, the voice and the orchestration that ties them together. These are the costs most often billed on top.

  • Telephony. Twilio lists US inbound calls to a local number at $0.0085 a minute plus $1.15 a month per number, and outbound calls at $0.0140 a minute. Retell charges $0.015 a minute on US local numbers and $0.06 on toll-free; on Vapi, carrier minutes are billed by the carrier.
  • The language model. Vapi passes model costs through at provider cost; Retell prices them per minute, from $0.0016 (GPT 5 nano) to $0.32 (GPT 6 Astra) on the standard tier, or $0.0032 to $0.16 for Claude models; its fast tier and speech-to-speech models cost more, up to $0.64 a minute. ElevenLabs bills them separately by model.
  • The voice. On Retell a platform voice is $0.015 a minute and an ElevenLabs v3 or v4 voice $0.10. Bought directly, ElevenLabs lists $0.04 to $0.08 per 1,000 characters and Deepgram Aura-2 $0.030.
  • Concurrency. Retell charges $8 a month per concurrent call above 20 and Vapi $10 a month per line above its package. With burst pricing enabled, ElevenLabs bills calls over the plan’s concurrency limit (up to 3x) at $0.16 a minute.
  • HIPAA. Vapi’s add-on is $2,000 a month. Elsewhere a BAA usually means an enterprise contract, and Twilio signs one only on its Security or Enterprise Edition (per its HIPAA page).

All-in cost per minute: a worked example

Stacking Retell AI’s own published components gives about $0.110 a minute, or about $552 a month at 5,000 minutes. This is our arithmetic, not a Retell quote. It assumes US calls on Retell’s Twilio numbers, a platform voice, Claude 4.5 Haiku, no add-ons, one phone number and no more than 20 calls at once.

Voice infrastructure             $0.055/min
Platform voice                   $0.015/min
Model: Claude 4.5 Haiku          $0.025/min
US telephony (Retell’s Twilio)   $0.015/min
Total per minute                 $0.110/min

5,000 min × $0.110 + $2 number = about $552 a month

The model moves the total most. Swap in Claude 5.5 Haiku at $0.0032 a minute and the same stack is about $0.0882 a minute, or about $443 a month, if it passes your call tests. Bundled rates such as Bland’s $0.14 already include the model and voice but not the call, so add your carrier’s rate before comparing. The AI voice agent cost guide runs the same sum for Vapi and a custom stack at 1,000, 10,000 and 50,000 minutes a month.

Platform or custom build?

For most first voice agents a platform is the cheaper start: no build to pay off, and the per-minute fee covers infrastructure you would otherwise run. A custom stack pays Twilio, Deepgram and the model vendor directly at the component prices above, and in our cost guide’s list-price model it starts paying back somewhere between about 11,000 and 31,000 minutes a month, depending on the platform it replaces. When we build one, it is a Fixed-Price Build of $25k–$120k, fixed after discovery; our AI voice agent development page covers what that includes and when we would tell you to stay on a platform.

How we check these prices

Once a month we open every vendor’s own pricing page or documentation and re-read each figure. The rules:

  • Vendor pages only. No review sites, directories or competitors’ comparison pages.
  • Logged-out list prices in US dollars, US rates where a region applies. Not negotiated quotes.
  • Each row keeps its source link and the date we last read it. A vendor with no published figure shows “No public price (sales quote)”.
  • Where a page contradicts itself, we say so in the row notes rather than pick a number.
  • No affiliate links, ratings or rankings.

Spotted a price that has changed, or a vendor we should add? Tell us and we recheck the source before the next update. Vendors can use the same route to correct their row.

Change log

    • First published: 11 voice agent platforms and 8 model, speech and telephony price lists, each read from the vendor’s own page.
    • Compared with our October 18, 2026 check for the AI voice agent cost guide, no Vapi, Retell, Twilio, Deepgram or Claude price changed.
    • Retell’s plan comparison lists HIPAA/BAA on pay-as-you-go as well as enterprise; we corrected the cost guide, which said enterprise only.

Sources

  1. Vapi pricing (checked October 19, 2026)
  2. Retell AI pricing (checked October 19, 2026)
  3. Bland AI pricing (checked October 19, 2026)
  4. ElevenLabs pricing (checked October 19, 2026)
  5. Deepgram pricing (checked October 19, 2026)
  6. Twilio pricing (checked October 19, 2026)
  7. Cartesia pricing (checked October 19, 2026)
  8. Amazon Web Services pricing (checked October 19, 2026)
  9. Google Cloud pricing (checked October 19, 2026)
  10. Synthflow pricing (checked October 19, 2026)
  11. PolyAI pricing (checked October 19, 2026)
  12. Twilio Programmable Voice (US) (checked October 19, 2026)
  13. ElevenLabs Text to Speech API (checked October 19, 2026)
  14. OpenAI Realtime API and GPT-Live (checked October 19, 2026)
  15. Google Gemini Live API (checked October 19, 2026)
  16. Anthropic Claude API (checked October 19, 2026)
  17. Twilio HIPAA (BAA requirement) (checked October 19, 2026)
FAQ

Pricing tracker questions.

How much does Retell AI cost per minute?

Retell lists $0.07 to $0.31 a minute for voice agents (checked October 19, 2026). The rate is built from parts: $0.055 a minute for voice infrastructure, $0.015 for a platform voice, the model (for example $0.025 for Claude 4.5 Haiku) and $0.015 for telephony on Retell’s US local Twilio numbers, which adds up to $0.110 a minute by our sum. Retell phone numbers cost $2 a month.

How much does Bland AI cost, and is there a monthly fee?

Bland’s Start plan costs $0.14 a talk minute with no platform fee. Its Build plan costs $0.12 a minute plus a $299 monthly platform fee, and Enterprise is priced to volume. The per-minute rate covers the model, speech-to-text and voice; telephony is billed separately at pass-through cost, and transfers cost $0.05 or $0.04 a minute (checked October 19, 2026).

Is Vapi cheaper than Retell?

Their headline rates measure different things, so they don’t compare directly. Vapi’s $0.05 a minute is hosting only, with speech, model and voice passed through at provider cost and carrier minutes billed by the carrier. Retell’s $0.055 is voice infrastructure, with voice, model and telephony priced per minute on top. Which costs less depends on the model and voice you choose.

Is telephony included in AI voice agent pricing?

Usually not. Vapi’s own telephony is free, but Twilio, Vonage and Telnyx minutes are billed by the carrier. Retell charges $0.015 a minute on US local numbers and $0.06 on toll-free, Bland passes carrier costs through, Cartesia charges $0.014 a minute on its numbers and Amazon Connect adds standard telephony rates. ElevenLabs adds no telephony fee, but your carrier bills you for its side of the call.

Which voice AI platforms offer HIPAA and a BAA, and what does it cost?

Vapi is the only platform here that publishes a price: a $2,000 a month HIPAA add-on. Retell’s plan table lists HIPAA and a BAA on pay-as-you-go and enterprise. Bland, ElevenLabs, Cartesia and Deepgram offer BAAs on enterprise plans at custom prices, and Twilio signs a BAA only on its Security or Enterprise Edition. Every vendor that handles patient data on the call needs to sign one.

How many concurrent calls do voice AI platforms include?

Retell includes 20 and charges $8 a month for each extra. Vapi includes 4, 10 or 30 by package and charges $10 a month per extra line. Bland allows 10 on Start and 50 on Build, ElevenLabs 4 to 40 by plan and Cartesia 8 to 60. Size it for your busiest hour, not your average.

How often is this pricing tracker updated, and where do the prices come from?

Monthly; the last full check was on October 19, 2026. Every price is read from the vendor’s own pricing page or documentation, logged out, in US dollars, and each row links its source and shows the date it was checked. No affiliate links, ratings or rankings. If a price has changed, tell us through the contact page and we recheck the source.