Voice AI Pricing in 2026: What Retell, Vapi, Bland and ElevenLabs Charge Per Minute, and One Bill Priced on All Four
As of October 8, 2026, Retell lists $0.07 to $0.31 a minute depending on model and voice, Vapi charges $0.05 a minute for hosting plus the transcriber, model and voice at cost, Bland charges $0.14 a minute all-in on its free Start plan, and ElevenLabs agents run $0.08 a minute. On the same 1,200-minute month the bills land between about $111 and $184, and concurrency limits, not headline rates, set the order.
What do voice AI platforms charge per minute? The short answer
As of October 8, 2026, Retell lists $0.07 to $0.31 a minute depending on the model and voice, Vapi charges $0.05 a minute for hosting plus the transcriber, model and voice at cost, Bland charges $0.14 a minute all-in on its free Start plan ($0.12 on Build, with a $299 monthly fee), and ElevenLabs agents work out to $0.08 a minute on every plan. On the same 1,200-minute month, the bills land between about $111 and $184. The headline rates are not comparable, because each one includes different parts of the call.
This page is for whoever is building the agent: the platform rate cards, read from each vendor's own pricing page on October 8, 2026, and one bill priced on all four. If you are a business owner pricing the whole project, setup and all, our AI voice agent cost guide covers that.
Why the headline per-minute rates mislead
A voice agent call has five moving parts: the platform that runs the conversation, speech-to-text, the language model, the voice, and the phone line. Each vendor's headline number covers a different subset of those:
- Vapi's $0.05 is hosting only. Transcriber, model and voice are passed through at the providers' cost, with no markup.
- Retell's $0.07 to $0.31 is its own estimate range. Behind it sit separate line items: $0.055 for its voice infrastructure, a voice from $0.015, a model from $0.0016, and $0.015 for telephony.
- Bland's $0.14 includes the model, transcription and voice in one rate, with no token charges. Telephony is billed separately.
- ElevenLabs' $0.08 covers the agent and voice. The model is billed on top, and your phone provider bills the carrier side.
Bland's own pricing page makes the same point about its rivals, and it is right that a platform-only rate understates the bill. It also cuts the other way: Bland's all-in rate is the highest headline of the four.
Free calculator
What are missed calls costing you?
Run the 60-second Missed-Call Cost Calculator — most service businesses find five figures a year leaking to voicemail.
Calculate my missed-call cost →Free, no email required to run it.
The rate cards, side by side
| Line item | Retell | Vapi | Bland | ElevenLabs agents |
|---|---|---|---|---|
| Platform | $0.055/min voice infra | $0.05/min hosting | Included in $0.14 (Start) or $0.12 (Build) | $0.08/min past included minutes |
| Speech-to-text | Included | Deepgram $0.0095 to $0.0099/min | Included | Included |
| Model | $0.0016 to $0.32/min (e.g. GPT 5.4 mini $0.024, Claude 4.5 Haiku $0.025) | OpenAI $0.0077 to $0.0452/min at cost | Included, no token charges | Billed separately (Gemini 2.5 Flash $0.0029/min) |
| Voice | $0.015/min (Retell, Cartesia, OpenAI voices) to $0.10 (ElevenLabs v3) | ElevenLabs $0.0146 to $0.0238/min at cost | Included | Included |
| Telephony | $0.015/min; no charge on your own SIP trunk | $0 on Vapi telephony or SIP in its estimate | Your carrier, or Bland's at pass-through cost | Your provider bills you; no ElevenLabs fee |
| Phone number | $2/mo | 1 included (5 on Core) | 1 inbound number included on Start | Your provider's |
| Monthly fee | None | None, or Core $29, or Pro from $999 | None on Start, $299 on Build | $0 to $990 by plan |
| Concurrent calls included | 20, then $8 per call a month | 4 (10 on Core), then $10 per line a month | 10 (Start), 50 (Build) | 4 to 40 by plan; burst at $0.16/min |
Every rate above is from the vendor's own page on October 8, 2026. Retell bills to the second and stops the agent fee once a call is transferred; Bland charges $0.05 a minute for transferred time on Start ($0.04 on Build).
A worked bill: 1,200 minutes on each platform
The workload: an inbound agent taking 400 calls a month at 3 minutes each, with a busy hour of about 8 calls at once. Where a platform needs your own phone line, we price a Twilio local number at $1.15 a month plus $0.0085 a minute to receive calls, which is $11.35 for this volume; our Twilio pricing page has the full rate card. This is our arithmetic on the published rates.
| Platform and setup | Math | Monthly bill |
|---|---|---|
| ElevenLabs Creator + Gemini 2.5 Flash + Twilio | $22 + 925 extra min × $0.08 + 1,200 × $0.0029 + $11.35 | about $111 |
| Vapi Core, lowest listed components | $29 + 1,200 × ($0.05 + $0.0095 + $0.0077 + $0.0146) | about $127 |
| Retell, own voices + GPT 5.4 mini + Retell telephony | 1,200 × ($0.055 + $0.015 + $0.024 + $0.015) + $2 number | about $133 |
| Bland Start + Twilio | 1,200 × $0.14 + $11.35 | about $179 |
| Retell, recommended GPT 5.6 Terra | 1,200 × ($0.055 + $0.015 + $0.064 + $0.015) + $2 | about $181 |
| Vapi Core, highest listed components | $29 + 1,200 × ($0.05 + $0.0099 + $0.0452 + $0.0238) | about $184 |
Three things decide the order, and none of them is the headline rate:
- Concurrency. Vapi's free tier allows 4 calls at once. An 8-call busy hour needs either Core at $29 (10 calls) or 4 extra lines at $10 each, so Core is in the bill above. Retell's 20 free concurrent calls cover this workload for nothing. On ElevenLabs, Creator's 10 slots fit; Starter's 6 would push 2 calls onto burst pricing at double the rate.
- The model. On Retell, switching from GPT 5.4 mini to the recommended GPT 5.6 Terra adds $48 a month at this volume. Model choice moves the bill more than platform choice does.
- Bundling. Bland is the simplest bill and the most expensive one here. You trade the ability to pick a cheaper model for one rate that never surprises you.
Our ElevenLabs pricing page explains why its plans all price minutes at about 8 cents, so you size them by concurrency instead.
When the bigger plans pay off
- Bland Build saves $0.02 a minute but costs $299 a month, so it only beats Start above 14,950 minutes a month. Most teams move up for the 50 concurrent calls and 2,000-calls-a-day cap, not the rate.
- Vapi Pro is 10% of your hosting fee with a $999 monthly minimum. At $0.05 a minute hosting, 10% only passes $999 at about 200,000 minutes a month, so below that you are paying the minimum for 30 concurrent calls, a 1-business-day support SLA and a 99% uptime SLA.
- Retell has no plan tiers below Enterprise. Growth costs $8 per extra concurrent call a month past 20.
- Bring your own phone line on Retell, and its $0.015 telephony charge and $2 number drop to zero. Twilio's $0.0085 a minute inbound plus $1.15 for the number comes to $11.35 instead of $20, about $9 a month saved at 1,200 minutes. Small, but it also keeps the number in your name if you change platforms.
Which voice AI platform should a builder pick?
On price alone, at small-business volume, the four land within about $75 of each other once concurrency is counted. The choice usually comes down to how much control you want. Vapi and Retell let you pick each component and swap models when a cheaper one ships. Bland trades that for one predictable rate. ElevenLabs is the voice-first option with plan-based concurrency. For latency, interruption handling, model support and developer experience, see Vapi vs Bland vs Retell.
The call is usually only the front of the job: once the agent captures a lead, a workflow still has to create the contact, text the caller and alert the team, and that runs on a second meter. Our automation platform pricing comparison prices that lead workflow on Zapier, Make, n8n, Power Automate and Lindy, from $12 a month at 300 leads on Make.
One option outside this table: Amazon Connect bills $0.038 a minute with Amazon's Lex bot and Nova Sonic speech model included, so the same 1,200 minutes come to about $49 with a US number (aws.amazon.com, read October 21, 2026), in exchange for building inside AWS. Our Amazon Connect pricing page has the full bill and the charges that surprise people.
Telnyx, a phone carrier, now sells its own agent platform too: $0.05 a minute with speech-to-text and voice included, the model billed per token (Telnyx estimates about $0.006 a minute for Kimi) and the phone line from $0.0052 a minute (telnyx.com, read October 23, 2026), so the same 1,200 minutes come to about $75 before its per-call rounding to the next full minute. Our Telnyx pricing page has the full rate card.
Microsoft shops have a fourth route: Copilot Studio prices voice in Copilot Credits, 35 a minute for generative voice with the agent's answers included, which is about $0.28 a minute when you use a full $200 pack of 25,000 credits (Microsoft Learn, read November 5, 2026). The 1,200 minutes above would come to 42,000 credits, or two packs. Our Copilot Studio pricing page covers the credit math.
LiveKit, the open-source framework many builders use underneath their own agents, prices the other way round: a flat $50 a month on its Ship plan covers the first 5,000 agent minutes, and the phone line ($0.01 a minute past 100) and the models are billed on top (livekit.io, read October 25, 2026). The same 1,200 minutes come to about $74 with a budget voice and about $110 with Cartesia's. Our LiveKit pricing page has the plan table and three model stacks.
What are the best Synthflow alternatives?
Synthflow's pricing page, read October 26, 2026, lists one option: Enterprise contracts starting at $30,000 a year, scoped around call volume, concurrency, telephony, integrations and security. That is a floor of $2,500 a month. At the Retell rate in the table above (about $0.109 a minute with GPT 5.4 mini and Retell's phone line), $2,500 buys about 23,000 minutes, roughly 7,600 three-minute calls a month. A small business taking 400 calls a month pays $111 to $184 on the platforms priced on this page.
- Closest replacement for a no-code builder: Retell or ElevenLabs. Both bill by the minute with no contract, and Retell includes 20 concurrent calls.
- One predictable rate: Bland, at $0.14 a minute on Start with no platform fee.
- Most control for a developer: Vapi, or LiveKit if you want to run the agent framework yourself.
- Cheapest per minute inside a cloud you already use: Amazon Connect at about $49 for the same 1,200 minutes.
If you still want a managed enterprise contract, ask each vendor to price your real minutes and busiest hour against Synthflow's $30,000 floor before you sign.
Who can build a voice AI agent for my business?
You can build on any of these platforms yourself if you can write the prompts, wire the calendar or CRM, and test call flows. Otherwise, hire a freelancer who knows one platform well, or a studio that builds and maintains the whole system. Ask any builder to price your agent on your real call volume and busiest hour, as above, before they pick the platform. To compare the studios and agencies that do this work, see our guide to choosing an AI automation agency.
SuperDupr builds AI voice agents for service businesses. Our published price for a voice agent build is $5,000 to $10,000, plus $200 to $2,000 a month ongoing depending on volume, and most AI builds go live in 2 to 4 weeks. The platform and phone accounts are set up in your business's name, so usage is billed at the vendors' own rates above. It starts with a free 30-minute strategy session.
Sources
- Retell AI pricing, retellai.com/pricing, read October 8, 2026: $0.07 to $0.31 range, voice infrastructure, voice, model and telephony rates, add-ons, phone number, concurrency and billing rules.
- Vapi pricing, vapi.ai/pricing, read October 8, 2026: $0.05 hosting, component ranges, packages (Core $29, Pro 10% with $999 minimum), concurrency, add-on lines, $5 free credits.
- Bland pricing, bland.ai/pricing, read October 8, 2026: Start $0.14 and Build $0.12 per minute, $299 Build fee, transfer rates, caps, concurrency, telephony policy.
- ElevenLabs agents pricing, elevenlabs.io/pricing/agents, read October 8, 2026: plan prices, included minutes, concurrency, $0.08 and $0.16 burst rates, Gemini 2.5 Flash example rate, telephony policy.
- Twilio Voice pricing, United States, twilio.com/en-us/voice/pricing/us, read October 8, 2026: local number and inbound minute rates.
- Synthflow pricing, synthflow.ai/pricing, read October 26, 2026: Enterprise only, contracts from $30,000 a year, scoped by call volume, concurrency, telephony, integrations, security and launch support; no self-serve plan listed.
- Worked bill: SuperDupr's arithmetic on the published rates above, for an assumed 400 calls of 3 minutes with 8 at once in the busiest hour.
Frequently Asked Questions
-
Retell lists $0.07 to $0.31 a minute as of October 8, 2026. The bill is built from line items: $0.055 a minute for its voice infrastructure, $0.015 for its own voices (up to $0.10 for premium ElevenLabs voices), the model from $0.0016 to $0.32, and $0.015 for telephony, which is free if you bring your own SIP trunk. Numbers are $2 a month and the first 20 concurrent calls are included.
-
Vapi charges $0.05 a minute for hosting, and passes the transcriber, model and voice through at the providers' cost: about $0.0095 for Deepgram, $0.0077 to $0.0452 for OpenAI and $0.0146 to $0.0238 for ElevenLabs. Vapi's own estimate for 1,000 minutes is $82 to $129. The free tier allows 4 concurrent calls; the Core package is $29 a month for 10.
-
As of October 8, 2026, Bland's Start plan is $0.14 a minute with no platform fee, and Build is $0.12 a minute plus $299 a month. Both rates include the model, transcription and voice, with no token charges. Transferred time is $0.05 a minute on Start and $0.04 on Build, and telephony is billed separately.
-
At 1,200 minutes a month with 8 calls at once in the busiest hour, our arithmetic on each vendor's published rates puts ElevenLabs Creator with a Twilio line at about $111, Vapi Core at $127 to $184 depending on components, Retell at $133 to $181 depending on the model, and Bland Start at about $179. The model you choose and the concurrency limit move the bill more than the platform does.
-
Most do once you pass the included amount. Retell includes 20 concurrent calls, then $8 per call a month. Vapi includes 4 on the free tier and 10 on Core, then $10 per line a month. Bland allows 10 on Start and 50 on Build. ElevenLabs ties concurrency to the plan (6 to 40) and charges burst pricing at $0.16 a minute above it.
-
As of October 26, 2026, Synthflow lists only Enterprise contracts starting at $30,000 a year, so small teams usually move to a pay-per-minute platform. Retell and ElevenLabs are the closest no-code replacements, Bland bills one $0.14 a minute rate on Start, and Vapi or LiveKit suit developers. On a 1,200-minute month those cost about $111 to $184, against Synthflow's $2,500 a month floor.
Written by
Justin McKelveyFormer CTO at an AI startup and former Head of Product, with 50+ products shipped. I ran snowboard and bike shops before I taught myself to code, so I build AI systems and websites with the owner in mind.