How Much Does an AI Voice Agent Cost Per Minute? About $0.07 to $0.31
By Sharrk Team · October 28, 2026
An AI voice agent costs about $0.07 to $0.31 a minute on the build-your-own platforms' published rates, before phone numbers and monthly fees. Sharrk is the all-in-one option: AI call minutes come out of the credits in your plan, and the same plan includes telephony, texting agents, follow-up Strikes, a CRM and Mako, the AI assistant that builds it all for you.
Below is where that range comes from, what each per-minute price includes, and how the all-in-one math compares. If you want monthly plan prices for AI receptionist products instead, see how much an AI receptionist costs.
How much does an AI voice agent cost per minute?
Most voice agent platforms bill by the minute of talk time, and the price you see on a pricing page is often only one layer of the bill. Here is what each vendor publishes on its own site, checked October 28, 2026.
| Option | Published per-minute price | Included | Billed on top |
|---|---|---|---|
| Sharrk (all-in-one) | 150 credits a minute (270 for premium voice/model); about $0.31 effective if all Core credits go to calls | Voice agent, telephony, texting agents, follow-up Strikes, CRM, Mako | Phone number: 10,000 credits a month, from the same credits |
| Retell AI | $0.07 to $0.31 a minute (pay as you go) | Voice infrastructure $0.055, voice from $0.015, LLM priced per minute | Telephony $0.015/min, numbers $2/month, add-ons |
| Vapi | $0.05 a minute hosting | Hosting only | Transcriber, LLM, voice at provider cost; telephony from the carrier |
| Bland AI | $0.14 a minute (Start), $0.12 (Build, $299/month) | LLM, speech-to-text, voices | Telephony, transfers $0.04 to $0.05/min |
| ElevenLabs Agents | $0.08 a minute ($0.16 burst) | Voice agent minutes | LLM "based on usage," telephony "at cost" |
| Synthflow | Not published per minute | Scoped by sales | Enterprise contracts start at $30,000 a year |
| OpenAI Realtime | Per token, not per minute | Speech-to-speech model | Telephony, hosting, everything else |
Where does the $0.07 to $0.31 range come from?
The range is built from the per-minute totals the vendors publish themselves, not from our own estimates:
- Low end, about $0.07 to $0.08. Retell's pricing page states "$0.07-$0.31 / min" for AI voice agents. Vapi's own calculator shows $82 to $129 for 1,000 minutes, which works out to about $0.08 to $0.13 a minute. ElevenLabs lists $0.08 a minute before the LLM.
- Middle, about $0.11 to $0.14. Retell's example estimate is $0.11 a minute ($0.055 infrastructure, $0.04 LLM, $0.015 voice). Bland's Start plan is $0.14 a minute plus telephony.
- High end, about $0.31. Retell's published top of range is $0.31. Pricier choices push past it: Retell lists speech-to-speech models such as GPT Realtime 2 at $0.38 a minute, and ElevenLabs charges $0.16 a minute for burst calls.
So "about $0.07 to $0.31" covers the standard setups on these pages. It does not include phone numbers, monthly platform fees, concurrency add-ons or the time it takes someone to build and maintain the agent.
What's included in a per-minute price, and what's extra?
A voice agent minute has five layers. Build-your-own platforms price them separately; all-in-one products bundle them.
| Layer | What it does | Example published rate |
|---|---|---|
| Platform or hosting | Runs the call and the agent logic | Vapi $0.05/min; Retell $0.055/min |
| Speech-to-text | Hears the caller | Vapi lists Deepgram at $0.0095 to $0.0099/min |
| LLM | Decides what to say | Retell prices each model per minute; Bland includes it; Vapi and ElevenLabs bill it on top |
| Voice | Speaks the reply | Retell $0.015/min standard, up to $0.10 for premium ElevenLabs voices |
| Telephony | Connects the phone call | Retell $0.015/min; Twilio via Vapi $0.008 in, $0.014 out |
Then come the monthly items: Retell numbers are $2 a month and extra concurrency is $8 per call slot a month; Vapi's Core plan is $29 a month; Bland's Build plan is $299 a month.
OpenAI's Realtime API is priced by tokens: $32 per million audio input tokens and $64 per million audio output tokens for GPT-Realtime-2 (May 7, 2026). OpenAI doesn't publish a per-minute figure for that model; its 2024 launch post estimated about $0.06 a minute of audio input and $0.24 a minute of audio output for the earlier preview model.
Is build-your-own or all-in-one cheaper per minute?
On the minute rate alone, build-your-own platforms usually look cheaper. The real comparison is what you get for the money:
- Build-your-own (Retell, Vapi, Bland, ElevenLabs) gives you a voice engine. You or a developer still connect the calendar, the CRM, texting, follow-up and reporting, and you pay each layer separately.
- All-in-one (Sharrk) prices minutes in credits that also cover texts, emails and voicemail drops, and the agent comes with the follow-up, booking and CRM already wired in.
For a business owner, the question is usually "what does a booked job cost?" rather than "what does a minute cost?" The same logic applies to human answering services; see answering service per minute vs per call.
How do you estimate your monthly voice agent bill?
Use this quick calculation for any platform:
- Count minutes. Calls per month × average call length. 300 calls at 3 minutes is 900 minutes.
- Price every layer. Platform + LLM + voice + telephony per minute, or one bundled rate.
- Add monthly fees. Phone numbers, plan fees, concurrency.
- Add build time. Hours to set up prompts, integrations and follow-up, and hours each month to change them.
How much does Sharrk cost per AI call minute?
Sharrk bills AI call minutes in credits: 150 credits a minute, or 270 for a premium voice or model. Here is what that means on each plan:
| Plan | Price | Credits | AI call minutes if all credits go to calls |
|---|---|---|---|
| Sharrk Free | $0 | 2,500 one-time | Build and test with Mako |
| Core | $99/month | 47,500 a month | About 316 |
| Complete | $449/month | 185,000 a month | About 1,233 |
The effective rate on Core, if all credits go to calls, is $99 ÷ 316 ≈ $0.31 a minute. A phone number uses 10,000 credits a month from the same balance, so with one number Core leaves 37,500 credits, about 250 AI call minutes. Complete also includes a done-for-you build every month. See Sharrk pricing for the full plan list.
That one rate covers far more than a minute of talk time:
- Voice agent that answers every call 24/7 and can live-transfer to a person, cold or warm.
- Telephony with your Sharrk business number.
- Texting agents that hold real conversations, plus replies to new leads in under 30 seconds by text, call or email.
- Follow-up Strikes: automatic AI calls, texts, emails, voicemail drops and review requests that branch on what the lead does.
- CRM for every lead, call and conversation.
- Booking onto Sharrk's calendar, Calendly, Cal.com or ServiceTitan.
Credits flex across channels, so a month with fewer calls leaves more for texts and follow-up.
How does Sharrk keep voice agent costs simple?
Sharrk gives you one plan and one credit balance instead of five vendor bills. Your AI voice agent answers calls, books jobs and hands off to Strikes without a developer wiring layers together.
Mako, Sharrk's AI assistant, does the building. Tell it in plain English, typed or spoken, what you want ("answer after-hours calls for Brightline Electric (sample), book estimates and text a confirmation") and it reads your website, drafts the agent, and sets up the Strikes and calendar types, and drafts your texting registration. Every change shows on an Apply card, Mako waits for your OK, and you can Undo.
Start on Sharrk Free with 2,500 credits to build and test with Mako, or book a demo to see the math for your call volume.
Related on Sharrk: AI Voice Agent · AI Receptionist · Speed to Lead · Pricing
Sources
- Retell AI, "AI Phone Agent Pricing": https://www.retellai.com/pricing (checked October 28, 2026)
- Vapi, "Pricing": https://vapi.ai/pricing (checked October 28, 2026)
- Bland AI, "Pricing": https://bland.ai/pricing (checked October 28, 2026)
- ElevenLabs, "ElevenAgents Pricing": https://elevenlabs.io/pricing/agents (checked October 28, 2026)
- Synthflow, "Pricing": https://synthflow.ai/pricing (checked October 28, 2026)
- OpenAI, "GPT-Realtime Model": https://developers.openai.com/api/docs/models/gpt-realtime (checked October 28, 2026)
- OpenAI, "Advancing voice intelligence with new models in the API" (May 7, 2026): https://openai.com/index/advancing-voice-intelligence-with-new-models-in-the-api/ (checked October 28, 2026)
- OpenAI, "Introducing the Realtime API" (October 1, 2024): https://openai.com/index/introducing-the-realtime-api/ (checked October 28, 2026)
Frequently asked questions
How much does an AI voice agent cost per minute?
About $0.07 to $0.31 a minute on the published rates of build-your-own platforms such as Retell, Vapi, Bland and ElevenLabs, checked October 28, 2026. Phone numbers, monthly plan fees and premium models are extra.
Does the per-minute price include the phone call?
Often not. Retell charges telephony at $0.015 a minute, Bland bills telephony separately, Vapi passes carrier rates through, and ElevenLabs lists telephony "at cost." All-in-one products like Sharrk include telephony in the plan.
Is the LLM included in voice agent pricing?
It depends on the platform. Bland says its rate has "no token charges," while ElevenLabs and Vapi bill the LLM on top, and Retell prices the LLM per minute by model.
How much does OpenAI's Realtime API cost per minute?
OpenAI prices GPT-Realtime-2 by tokens, not minutes: $32 per million audio input tokens and $64 per million audio output tokens. Its 2024 launch post estimated about $0.06 a minute of input and $0.24 a minute of output for the earlier preview model.
How much does Sharrk cost per AI call minute?
Sharrk uses 150 credits per AI call minute (270 for a premium voice or model). Core is $99 a month with 47,500 credits, about 316 AI call minutes or roughly $0.31 a minute if all credits go to calls, and that includes telephony, texting agents, Strikes, CRM and Mako.