Vapi vs Retell AI vs Bland AI vs Synthflow (2026): Real Cost per Minute for a Small Business

Vapi, Retell AI, Bland AI and Synthflow are the four voice AI platforms that come up when a small business wants a phone agent that answers calls, qualifies callers and books appointments. None of them is a giant: they are young, fast-moving companies, and their pricing pages changed a lot in 2026. That is exactly why most comparisons online are already out of date. This guide uses only the figures published on each vendor's own pricing page, read on November 15, 2026, and turns them into what you actually pay for 1,000 minutes of calls a month, which is roughly 330 calls of three minutes for a busy clinic, law office or trades business.

At a glance

Every figure in this table comes from vapi.ai/pricing, retellai.com/pricing, bland.ai/pricing and synthflow.ai/pricing on November 15, 2026. Voice AI prices move every few months, so check the page the day you sign.

Why per-minute prices are hard to compare

A voice agent is not one product. Every second of a call runs through four pieces: speech-to-text (the transcriber), a language model that decides what to say, text-to-speech (the voice), and telephony (the phone line). Each platform bundles these differently.

• Vapi charges a hosting fee of $0.05 per minute and passes every provider through at cost, so your bill depends on the transcriber, model and voice you pick.
• Retell AI charges a voice infrastructure fee of $0.055 per minute and prices each model, voice and phone line as its own line, all visible in its calculator.
• Bland AI folds the model, transcription and voice into one rate per minute. Telephony is separate, on your own carrier or Bland's at pass-through cost.
• Synthflow does not publish a rate at all.

So a "$0.05 per minute" headline and a "$0.14 per minute" headline are not comparable until you add the missing pieces. Bland makes the same point on its own pricing page, where it estimates complete stacks at roughly $0.13 to $0.30 per minute on Vapi and $0.11 to $0.25 per minute on Retell. That is a competitor's estimate, so below we rebuild the numbers ourselves from each vendor's published rates.

What 1,000 minutes a month really costs

Vapi

Vapi's own calculator, set to 1,000 minutes, shows:

• Hosting: $50 (1,000 x $0.05)
• Transcriber (Deepgram, $0.0095 to $0.0099 per minute): about $10
• Model (OpenAI, $0.0077 to $0.0452 per minute): $8 to $45
• Voice (ElevenLabs, $0.0146 to $0.0238 per minute): $15 to $24
• Vapi telephony or SIP: free. On Twilio instead, add $0.008 per minute inbound, so about $8.

Total: $82 to $129 a month, the range Vapi itself displays, plus the optional Core package at $29 a month if you want 10 concurrent calls, 5 phone numbers, zero data retention and an email support SLA. Without a package you get 4 concurrent calls and community support on Discord.

Retell AI

Retell lists every component. For a US line on Twilio at $0.015 per minute, two realistic builds:

Two billing rules matter for small businesses. A call under 10 seconds with a dynamic greeting is billed as 10 seconds, so hang-ups and wrong numbers are not free. And if your prompt (instructions, tools, transcript and retrieved knowledge) averages over 4,000 tokens, billed time grows in proportion: a 4,800-token prompt bills 20% more time. Keep the prompt tight and the knowledge base focused. Pay as you go includes 20 concurrent calls, which is more than most small offices will ever need.

Bland AI

On the Start plan, every minute costs $0.14 with the model, transcription and voice included, and there is no platform fee. Transfers to a human cost $0.05 per minute. For 1,000 minutes: $140 a month plus telephony, which is either your own carrier or Bland's Twilio at pass-through cost.

The Start plan has real limits: 10 concurrent calls, 100 calls a day, 100 calls an hour, one voice and 10 knowledge bases. A practice that receives 330 calls a month stays far below them, but a Monday-morning rush of 120 calls would hit the daily cap.

The Build plan lowers the rate to $0.12 per minute but adds a $299 monthly fee. For the Build plan to pay off, the two cents saved per minute must cover $299, which happens at about 14,950 minutes a month. Below that, stay on Start.

Synthflow

Synthflow's pricing page now says it is sold as an annual enterprise agreement, scoped to your call volume, phone system, integrations, security needs and launch support, with no free plan, no trial and no rate by the minute. That may suit a contact center. For a ten-person business that wants to test an agent this week, it removes Synthflow from the shortlist.

Side by side for a small business

The cheapest platform on paper is not automatically the cheapest to run. Vapi and Retell reward someone who will test models and voices and watch the invoice. Bland costs more per minute but removes that work entirely. For a business owner without a technical person, the hours saved usually matter more than $40 a month.

Which one fits which business

Choose Retell AI if

You want low, predictable costs and a calculator you can show your accountant. Twenty concurrent calls on pay as you go, $10 of free credits, and every option priced to the cent make it the easiest platform to budget. It also integrates with n8n, Make, HubSpot and GoHighLevel according to its integrations page, which matters once the agent must book, log and follow up.

Choose Vapi if

A developer is building the agent and wants to pick every provider, swap models freely and pay them at cost. Vapi's catalog and flexibility are its strengths. The trade-off is four lines to watch on every invoice and only 4 concurrent calls until you buy a package or add lines at $10 each per month.

Choose Bland AI if

You want one number per minute and no surprise model bills. Bland states "no token charges" and "no model-provider pass-throughs" on its pricing page. It is the simplest budget conversation of the four, as long as your volume fits the Start plan's daily and hourly caps.

Choose Synthflow if

You run a large call operation with a procurement team, need contract terms, and are happy to commit annually.

What the platform does not do for you

Every platform above gives you a voice that can talk. None of them, on its own, knows your calendar rules, your service area, your prices, or what to do with the caller after the call. That is where most small-business voice projects succeed or fail:

1. 1 Booking. The agent must check real availability and create the appointment in your actual calendar or practice software, not a demo calendar.
2. 2 Logging. Every call should land in your CRM with the caller, the reason, the outcome and the recording link, so nobody calls back blind.
3. 3 Follow-up. Missed or unfinished calls need a text or email within minutes, and a human task when the agent could not help.
4. 4 Handoff. Urgent or sensitive calls must transfer to a person with the context already written down.

These are workflow problems, not voice problems. They are usually built in an automation tool such as n8n that receives each call's webhook, writes to the CRM and sends the follow-ups. If you are still deciding whether you even need a custom voice agent, compare it with the turnkey AI receptionist apps on our AI receptionist cost calculator: for many small offices a ready-made receptionist at a flat monthly price is the faster first step.

How to test before you commit

1. 1 Write down your ten most common call types, with the answer or action each one needs.
2. 2 Build the same simple agent on two platforms using their free credits ($5 on Vapi, $10 on Retell, 2 credits on Bland).
3. 3 Call each one twenty times, including with background noise, interruptions and an accent different from yours.
4. 4 Read every transcript. Count wrong answers, awkward pauses and failed bookings.
5. 5 Price your real monthly minutes with each vendor's calculator, including telephony.
6. 6 Only then connect the winner to your calendar and CRM.

Common mistakes

• Comparing headline rates without adding model, voice and telephony.
• Choosing the most expensive voice and model for a call that only needs to take a name and a time.
• Letting the prompt grow past 4,000 tokens on Retell, which silently raises billed minutes.
• Forgetting concurrency: 4 lines on Vapi without a package is tight for a Monday rush.
• Launching without a human fallback for urgent calls.

Frequently asked questions

Which voice AI platform is cheapest for a small business?

On published rates for 1,000 minutes a month, Vapi's own estimate of $82 to $129 and a lean Retell build at about $98 are the lowest. Bland is $140 plus telephony but includes the model and voice in one rate. The real cheapest option is the one your team can run without spending hours tuning it.

Is Vapi or Retell better?

Retell is easier to budget and includes 20 concurrent calls on pay as you go. Vapi gives a developer more freedom over providers at cost. If nobody on your team writes code, Retell or Bland will be simpler.

Does Synthflow have a free trial?

Not anymore. Its pricing page, read on November 15, 2026, says it has no free plan or trial and is sold as an annual enterprise agreement.

Can a voice agent book appointments directly?

Yes, all four can call tools during a call. Retell and Bland list appointment booking features on their sites. In practice you still need to connect your real calendar and decide the rules, which is where an automation layer comes in.

How long does it take to launch a phone agent?

Bland says most teams deploy a first agent within a day on its self-serve plans. A production agent connected to your calendar, CRM and follow-ups usually takes one to two weeks of building and testing.

If you want a phone agent that answers, books and logs every call in your own accounts, we build them on these platforms with n8n at a fixed price. Book a free automation audit and we will tell you which platform fits your call volume, even if the answer is a simpler tool.