Voice AI Pricing Compared: What You Actually Pay For

Voice AI pricing comes in four main models: a bundled rate per minute, itemized per-minute components, a fee per resolved outcome, and per-seat contact center licenses with AI sold as an add-on. The headline numbers are hard to compare because each model leaves different things out, so the useful comparison is what a month of your own calls would cost and what you get for it.

This page explains each model using prices the vendors publish on their own pricing pages, checked on October 13, 2026. Prices change often, so confirm current rates with each vendor before deciding.

What are the main voice AI pricing models?

Each model ties the bill to a different unit and leaves a different set of costs outside the headline rate.

• Bundled per minute: What you pay for: One rate for model, speech and voice. Often outside the headline: Telephony, platform fee, transfer minutes.
• Itemized per minute: What you pay for: Each component listed separately. Often outside the headline: Depends on vendor: providers, telephony, support.
• Per outcome: What you pay for: Each resolution the vendor counts. Often outside the headline: Minimums, voice priced higher.
• Per seat (CCaaS): What you pay for: Each human agent license. Often outside the headline: AI features, tokens, usage.

Many vendors mix models. A per-minute platform may add a monthly platform fee. A per-seat contact center suite may meter AI features in tokens. An enterprise contract may set a minimum annual spend on top of usage. The explainer on usage-based pricing covers the billing mechanics that apply to all of them.

What does a bundled per-minute rate include?

A bundled rate puts the language model, speech recognition and voice into one number per minute. Bland is a clear example. Its pricing page lists $0.14 per minute with no platform fee on the Start plan, and $0.12 per minute plus a $299 monthly platform fee on the Build plan. Bland says that rate covers the LLM, speech-to-text and text-to-speech, with "no token charges." Telephony is billed separately, either on your own carrier or on Bland's at pass-through cost. Transferred time has its own rate: $0.05 per minute on Start and $0.04 on Build.

Bundled rates are easy to forecast. The trade-off is less control over cost per call, since the model and voice behind the rate are not priced separately.

How does itemized component pricing work?

Itemized pricing lists each part of the call separately, so you can see which choice drives cost. There are two common versions.

Platform fee plus pass-through providers. Vapi charges $0.05 per minute for hosting and passes model, transcription and voice costs through "at cost." Phone transport from carriers such as Twilio is "charged by provider, not Vapi." Vapi's own calculator estimates $82 to $129 for 1,000 minutes with Deepgram transcription, an OpenAI model, an ElevenLabs voice and Vapi's free telephony option. Support packages sit on top: Core is $29 per month, and Pro is 10% of the hosting fee with a $999 monthly minimum. Vapi lists HIPAA as a $2,000 per month add-on and extra concurrency at $10 per line per month.

Every component billed by one platform. Retell AI lists each component on the Retell pricing page and bills them together. Voice infrastructure is $0.055 per minute. LLMs range from under a cent to $0.32 per minute on the standard tier. Voices run from $0.015 to $0.10 per minute, and telephony is $0.015 per minute on a Retell US number, with no Retell charge over your own SIP trunk. Pay-as-you-go has no platform fee. A common setup of GPT 5.1, a Retell platform voice and Retell telephony comes to $0.125 per minute.

The model you pick moves the per-minute rate more than anything else. On itemized pricing, LLM rates on the same page range from under a cent to $0.32 per minute, so price the exact model and voice you plan to run.

Retell's advertised range is $0.07 to $0.31 per minute. Its docs describe that range as voice infrastructure, text-to-speech and the LLM together, so telephony and add-ons are extra. Larger models on the same page, and fast tiers, can exceed the range on their own. The monthly cost guide builds a full estimate from these components.

What do enterprise contracts and quote-only pricing add?

Several vendors publish no per-minute rate at all, or publish only a starting contract size:

• Synthflow: its pricing page states that "Enterprise contracts start at $30,000 annually," with final pricing scoped around call volume, concurrency, telephony, integrations, security and launch support.
• PolyAI: its pricing page says ongoing use "is priced on a per-minute basis, which includes proactive performance improvements, maintenance and 24/7 support." No rate is published, so pricing is by quote.
• Retell Enterprise: custom pricing with volume rates, plus options such as a dedicated server and dedicated implementation support as add-ons.

With a contract minimum, the real per-minute cost depends on how much of the commitment you use. A minimum sized for a full rollout costs a lot per minute during a pilot that only runs one use case.

How is outcome-based pricing different?

Outcome-based pricing charges per resolved conversation or completed task instead of per minute. Sierra's site says customers "only pay for the value Sierra delivers with outcome-based pricing" but publishes no rate. Intercom publishes one for its Fin AI agent: $0.99 per outcome, with a 50-outcome monthly minimum when used with another helpdesk. Voice costs more: Intercom says "Fin Voice is available via sales at $1.99 per voice outcome."

The appeal is that you pay only when something works. Two questions decide whether that holds. First, who defines an outcome? Intercom counts a resolution when "no further help is requested after the last AI answer," and its billable outcomes also include a configured handoff to a human. Neither is the same as the customer's issue being solved. Second, how does an outcome compare with the minutes it used? A resolved issue costs the same whether the conversation took two minutes or ten, so short, simple resolutions can cost more than the minutes they used. You also pay a share of results that your own data, policies and systems helped produce. The blog post on usage-based vs. outcome-based pricing covers this trade-off in detail.

Where do per-seat contact center platforms fit?

Contact center as a service (CCaaS) suites price per human agent seat, and AI is layered on. Genesys Cloud CX 1 is listed at $75 per user per month, billed annually, and Genesys notes that AI Experience tokens "are required for certain features." Talkdesk lists Digital Essentials at $85 per user per month.

Per-seat pricing fits the human side of your operation. It is not a like-for-like comparison with voice agent pricing, because the seat pays for a person's tools, while voice agent pricing pays for the conversation itself. Many teams run both: a CCaaS platform for human agents and a voice agent platform that answers first and transfers into it, for example an AI answering service that covers after-hours and overflow calls before routing to the CCaaS queue.

How do you compare voice AI quotes on equal terms?

Turn every quote into the same two numbers: monthly cost for your expected volume, and cost per resolved call. Then check what each price leaves out.

• Is telephony included, and what does it cost after a transfer?
• Is there a platform fee, support package or annual minimum?
• Which LLM and voice does the quoted rate assume?
• How are calls rounded, and are silence and voicemail billed?
• How many concurrent calls are included, and what does more cost?
• Are compliance features such as HIPAA or a BAA priced separately?
• What counts as a billable outcome, if pricing is per outcome?

Volume changes which quote is cheaper. Take two hypothetical quotes. Quote A is $0.10 per minute, plus $0.01 per minute for telephony and a $500 monthly platform fee. Quote B is $0.15 per minute with telephony included and no platform fee. At 10,000 minutes a month, A costs $1,600 and B costs $1,500. At 50,000 minutes, A costs $6,000 and B costs $7,500. The two cross at 12,500 minutes, so run every quote at your low, expected and peak volumes.

Cost per resolution matters more than cost per minute. A cheaper agent that transfers more callers costs more overall, because each transfer adds a person's time. The cost per resolution benchmarks show the math, and the guide to evaluating a voice AI platform before going live covers what to test beyond price.

The most reliable comparison is your own calls on each platform. If you are evaluating vendors for a contact center rollout, ask the Retell team for a pilot on real calls before you sign anything.