AI Voice Agent Cost in 2026: Real Per-Minute Math

Every number below comes from the vendor's own pricing page as of October 2026. Prices move every few months, so treat this as a method you can rerun.

How much does an AI voice agent cost per minute?

It is a range, because five separate meters tick at the same time:

• Low end, about $0.09 per minute: a small, fast language model, a standard voice and a local phone number.
• Typical, about $0.11 per minute: Retell's own estimator example, and the middle of Vapi's calculator range.
• High end, $0.14 to $0.31 per minute: a large model, a premium voice, and add-ons like denoising or PII removal.

The single biggest swing is the language model. On Vapi's calculator, OpenAI model costs range from $0.0077 to $0.0452 per minute. That is a sixfold difference on one line, and it decides whether you sit at the top or bottom of the range.

What are the five meters on every AI phone call?

The caller speaks, the audio is transcribed, a model decides what to say, a voice speaks the reply, and a carrier moves the audio both ways. Each step bills separately.

Add the Vapi rows and you get $0.0818 to $0.1289 per minute before the phone line. That matches Vapi's own 1,000-minute example of $82 to $129 a month. Add Twilio's local inbound rate and the all-in figure becomes about $0.090 to $0.137 per minute.

A note on phone rates: Vapi's page lists Twilio inbound at $0.008 per minute, while Twilio's own US page says $0.0085. We use Twilio's figure, because that is what Twilio bills you.

Is Vapi cheaper than Retell or ElevenLabs Agents?

The three platforms price the same call in three different ways.

Vapi charges a $0.05 per minute hosting fee and passes the model, voice and transcription costs through at cost. You pick each component. The free tier allows 4 concurrent calls and one phone number. Extra concurrency is $10 per line per month, and the Core plan is $29 a month with 10 concurrent calls.

Retell AI quotes $0.07 to $0.31 per minute on pay-as-you-go. Voice infrastructure is $0.055 per minute, standard TTS $0.015, and telephony about $0.015. Its estimator example lands at $0.11 per minute. Retell bills to the nearest second, does not bill failed calls, gives you the first 20 concurrent calls free and then charges $8 per concurrent call per month. Numbers are $2 a month.

ElevenLabs Agents sells minute bundles. Pro is $99 a month with 1,238 minutes included, Business is $990 a month with 12,375 minutes. Extra minutes cost $0.08 on every plan, or $0.16 for burst usage. The language model and telephony are billed at cost on top.

So which is cheapest? With a small model, Vapi usually wins. With a large model and premium voice, all three land close together, and concurrency limits, voice quality in your callers' accents and tool support matter more than the rate.

What does an AI voice agent cost at 1,000, 5,000 and 10,000 minutes a month?

Here is the same workload priced three ways. Assumptions, so you can check the math:

• US local number on Twilio for Vapi and ElevenLabs ($0.0085/min inbound plus $1.15 a month for the number).
• Retell at its $0.11 per minute estimator example, plus one $2 number.
• ElevenLabs on the Pro plan, with the same OpenAI model range Vapi's calculator uses ($0.0077 to $0.0452 per minute) for the LLM line.
• Inbound calls only, no add-ons, within free concurrency limits.

Worked example: 1,000 minutes on Vapi

• Hosting: 1,000 × $0.05 = $50.00
• Speech-to-text: 1,000 × $0.0095 = $9.50
• Language model (small): 1,000 × $0.0077 = $7.70
• Voice: 1,000 × $0.0146 = $14.60
• Twilio inbound: 1,000 × $0.0085 = $8.50
• Twilio number: $1.15
• Total: $91.45 a month, or about $0.09 per minute

Swap in the large model ($0.0452) and the top voice rate ($0.0238) and the same month costs $138.55.

Worked example: 5,000 minutes on ElevenLabs Pro

• Plan: $99 with 1,238 minutes included
• Overage: 3,762 × $0.08 = $300.96
• Language model: 5,000 × $0.0077 = $38.50 (or $226 with the large model)
• Twilio inbound: 5,000 × $0.0085 = $42.50, plus $1.15 for the number
• Total: about $482 a month with a small model, $670 with a large one

At 10,000 minutes, ElevenLabs Pro with overage ($800 before LLM and phone) is still cheaper than the $990 Business plan. Business only pays off as you approach its 12,375 included minutes.

For context, CloudTalk's 2026 pricing guide to AI answering services puts a typical small business at $50 to $200 a month, with minute bundles at roughly $0.15 to $0.50 per minute. That is a vendor-published figure, but it lines up with the 1,000-minute row above. Packaged AI receptionist products usually charge more per minute in exchange for less setup.

Which AI voice agent costs do people forget?

The per-minute rate is the number everyone compares. These are the lines that surprise people on the first invoice.

Silence and hold time. Retell states plainly that it charges for silence and hold time. Most platforms bill connected minutes, not spoken ones. An agent that waits 20 seconds for a caller to find their card number is billing for those 20 seconds.

Concurrency. If six people call your restaurant at 6:30 pm and your plan allows four concurrent calls, two hear a busy signal or a fallback. On Vapi that fix is $10 per extra line per month. On Retell, the first 20 are free, then $8 each.

Phone numbers. Small, but recurring: $1.15 a month for a Twilio local number, $2 a month on Retell. Toll-free numbers cost more per minute too: Twilio US toll-free inbound is $0.0220 against $0.0085 for local.

Add-ons. Retell charges $0.01 per minute for PII removal, $0.005 for safety guardrails and $0.005 for advanced denoising. Each looks tiny. Stack all three and you have added $0.02 per minute, close to 20% of a typical bill.

Compliance. Vapi's HIPAA add-on is $2,000 a month, including a Business Associate Agreement. For a US clinic doing 1,000 minutes a month, that one line costs more than 14 times the call minutes. If you handle patient health information over voice, price this first. Our guide to what an AI receptionist does in a dental clinic covers where that data should live.

What does it cost to build an AI voice agent versus run it?

Running cost is the meter. Build cost is the work of making the agent useful: writing the prompts, connecting the tools it calls, setting handover rules, and testing it against real callers.

A demo that answers FAQs takes an afternoon. An agent that books into your real calendar, takes a 14-item order with modifiers, or knows when to hand a distressed patient to a human takes weeks. That second kind is worth paying for.

Shorai Studios, an AI studio that builds voice agents, WhatsApp agents and automations for small businesses, quotes AI solutions from $1,000, with a focused agent usually live in four to eight weeks. The quote is fixed and written after a free call. Running costs are billed by the platforms directly to accounts you own, so you see every meter yourself and nothing is marked up.

How can you cut per-minute cost without making the agent worse?

1. Route by difficulty. Use a small model for routine turns like greetings, hours and booking confirmation, and call a larger model only when the conversation gets hard. Since the LLM line swings sixfold, this is the biggest lever.
2. Keep prompts tight. Every turn sends the system prompt again. A 3,000-word prompt full of edge cases costs more on every turn than a short prompt plus a lookup tool.
3. Set a silence timeout. If the caller goes quiet for 10 to 15 seconds, prompt once, then end the call politely and send a text follow-up.
4. Transfer, don't hold. If a human needs to take over, transfer the call. Leaving the agent on the line while a caller waits on hold bills every second.
5. Use a local number. On Twilio US, local inbound is less than half the per-minute rate of toll-free.
6. Test cheaper voices. ElevenLabs on Vapi runs $0.0146 to $0.0238 per minute. Many cheaper voices sound fine.
7. Shorten the call. An agent that confirms an order in one clear readback instead of three saves 20 to 30 seconds per call. Across 1,667 calls, that is 550 to 830 minutes.

FAQ

How much does an AI voice agent cost per minute?

As of October 2026, a typical AI voice agent costs about $0.09 to $0.14 per minute all in, covering platform fee, speech-to-text, the language model, the synthetic voice and the phone line. Retell's own estimator example lands at $0.11 per minute. Premium voices and large models can push some setups past $0.30 per minute, while small models keep simple calls near the bottom of the range.

How much does an AI receptionist cost per month for a small business?

For a small business taking 1,000 minutes of calls a month, roughly 330 three-minute calls, expect about $90 to $155 a month in running costs as of October 2026, depending on platform and model. Vendor-published guides such as CloudTalk's put typical small business plans at $50 to $200 a month. A one-time build cost sits on top if you have the agent built for you.

Is Vapi cheaper than Retell?

It depends on the model you run. As of October 2026, Vapi's own 1,000-minute example totals $82 to $129 before phone line costs, while Retell's estimator example is $0.11 per minute, or $110 for 1,000 minutes. With a small model Vapi usually comes out cheaper. With a large model and premium voice the two land close together, so compare on concurrency and features too.

Do you pay for silence and hold time on AI calls?

Yes, usually. Retell states on its pricing page that it bills to the nearest second and charges for silence and hold time, while failed calls are not billed. Most platforms bill connected minutes, not spoken minutes. That is why silence timeouts, clean call endings and transferring callers instead of holding them on the agent's line matter for your monthly bill.

Is an AI receptionist cheaper than a human receptionist?

For answering calls, almost always yes. A 5,000-minute month is about 83 hours of talk time and costs roughly $450 to $690 to run on the platforms in this guide. The better framing is coverage: the agent answers every call at once, at night and on weekends, while your staff handle the conversations that need judgment. Most clinics and restaurants keep their people and stop missing calls.

What does HIPAA compliance add to voice AI cost?

On Vapi, the HIPAA compliance add-on is $2,000 a month and includes a Business Associate Agreement, as of October 2026. Retell offers a PII removal add-on at $0.01 per minute and safety guardrails at $0.005 per minute. For a small US clinic, the compliance line can cost more than the call minutes themselves, so price it before you pick a platform.

How long does it take to build an AI voice agent?

A focused voice agent that answers questions, books appointments or takes orders is usually live in four to eight weeks. The first working version can be ready much sooner. Most of the time goes into connecting the calendar, POS or CRM, writing handover rules, and testing with real calls in your accent mix before the agent goes on your main number.

Want a real number for your own calls?

Send us your monthly call volume and what callers usually want, and we will price the meters for your setup. See how we approach AI voice agent development for small businesses, including handover rules and data ownership.

For working examples, look at our AI phone ordering system for a Canadian restaurant brand, where every order from phone, web and chat lands in one kitchen queue, and Dentistro, an AI-first dental clinic front desk that books and reschedules over calls and WhatsApp. If you run a restaurant, our walkthrough of how AI phone ordering for restaurants works prices calls per order.

Your first version is on us: we build the base of your agent for free so you can hear it answer your own calls before you pay anything.

Sources

• Vapi pricing
• Retell AI pricing
• ElevenLabs Agents pricing
• Twilio Voice pricing, United States
• CloudTalk: How much does an AI answering service cost?