The Start tier carries no platform fee and no card requirement, so you can build and test an agent without a subscription. Calls still bill by the connected minute, and the cap of 100 calls a day with 10 concurrent lines limits a test. Free here means no fixed fee, not free calling.

Does Bland AI charge per minute or per call?

Both, in effect. Conversation time bills by the minute, prorated to the exact second, so a 40-second call is not rounded up. Every outbound attempt and failed call routed through Bland's own telephony then carries a small minimum charge, and voicemail bills as ordinary call time.

Can Bland AI be self-hosted?

On the Enterprise tier only. The pricing page lists on-prem and VPC deployment as an Enterprise item alongside a dedicated orchestration server, and Bland's enterprise docs describe isolated infrastructure dedicated to each enterprise customer. Isolation is documented as an enterprise feature, so an on-premise requirement moves the conversation to a contract.

Which is cheaper, Bland AI or Vapi?

It depends on your model and voice choices. Bland bundles the model, transcription and speech output into one rate; Vapi charges a lower hosting rate and passes those costs through. Price both against your own minutes, then add fixed fees. Vapi's pricing page includes a minutes calculator, so test it with your own volume.

What happens to the bill when nobody picks up?

A minimum charge still applies to each outbound attempt and to failed calls placed through Bland's telephony. It is small per attempt and adds up across a campaign, since most dials never connect. On your own carrier the minimum does not apply, but you pay that carrier's rates, so price them separately.

Bland AI Pricing and Plans Explained (2026)

Last updated: 1 October 2026

Quick Answer: Bland AI pricing is a bundled rate per connected call minute, about 14% lower on the paid Build tier than on Start. A monthly platform fee buys that lower rate, along with higher concurrency and daily caps. Transfers, unanswered attempts and SMS bill on top, and Enterprise is quoted against your call volume.

Bland is built for outbound calling at volume, which makes its billing model matter more than most.

A dialer also spends on attempts nobody answers and on calls passed to a human.

How much does Bland AI cost in 2026?

Bland charges for each connected minute at a rate set by your plan, and the rate includes the language model, transcription and speech output rather than passing each through at cost.

Bland's pricing page, read 1 October 2026, lists three tiers. Start has no platform fee, at $0.14 a connected minute with $0.05 transfers, and sign-up includes two credits and an inbound number Bland values at $15 a month. Build costs $299 a month and cuts the rate to $0.12, with $0.04 transfers. Enterprise is priced to contracted volume.

A Bland AI phone agent bills by the connected second, so a lower unit rate means a higher tier.

Plan	Platform fee (USD/month)	Connected minute (USD)	Transfer minute (USD)	Concurrent calls	Daily cap	On-prem or VPC
Start	0	0.14	0.05	10	100 calls	No
Build	299	0.12	0.04	50	2,000 calls	No
Scale	499	0.11	0.03	100	5,000 calls	Not documented
Enterprise	Contracted	Custom	Custom	Sized to volume	Unlimited	Available

Plan figures as documented by Bland on 1 October 2026. Scale appears only in Bland's billing documentation.

Bland AI plans compared: Start, Build, Scale and Enterprise rates and limits

What do Bland AI's plans include?

Each tier bundles a rate with limits, and the limits bite first: a daily cap rejects calls rather than overcharging.

Bland's billing documentation adds a Scale tier at $499 a month, with a $0.11 connected-minute rate and $0.03 transfers. The public pricing page omits it, so confirm it in the dashboard before budgeting on it. It also sets a $0.015 minimum charge on every outbound attempt and failed call through Bland's telephony, SMS at $0.02 a message both ways, widget messages at $0.01 each, and voicemail billed as ordinary call time.

Bland AI competitors usually split the same work into separate charges for the model, the voice and the carrier, which is why two quotes with similar headline numbers land far apart. Two details move budgets more than the rate. Call time is prorated to the second, and transfers are free to customers who bring their own Twilio account. Concurrency, not the rate, sets throughput: ten lines cannot run a 5,000-dial day.

What does Bland AI enterprise pricing cover?

Bland AI enterprise pricing is a contract rather than a listed rate, and what it adds is isolation and controls.

The pricing page puts concurrency sized to your volume, on-prem or VPC deployment, a forward deployed engineer, a BAA, SSO and data residency on that tier, along with warm and live transfers, guardrails and the dedicated orchestration server. Bland's enterprise documentation says each enterprise customer runs on its own isolated infrastructure, and its pricing FAQ states that it holds SOC 2 Type I and Type II, HIPAA eligibility with a signed BAA, GDPR and PCI DSS.

If a security review requires SSO, a BAA or data residency, the self-serve rate is not the rate you will pay.

What does an outbound campaign cost on Bland AI?

A campaign's real unit cost is the monthly total divided by connected calls, which at typical outbound answer rates runs close to double the headline rate.

Take a month of 5,000 dial attempts on the Build tier, a 30% connect rate, an average connected call of 2.4 minutes, and one call in five transferred to a human for two minutes. Rates are Bland's Build-tier rates read 1 October 2026; the totals are our own arithmetic on them, not a Bland quote, and assume Bland's own telephony.

Line item	This campaign	Subtotal (USD)
Platform fee	Build tier, one month	299.00
Connected minutes	3,600 minutes at the Build rate	432.00
Transfer minutes	600 transfer minutes at the Build transfer rate	24.00
Unanswered attempts	3,500 attempts at the minimum charge	52.50
Monthly total		807.50

Divide that total by the 1,500 calls that connected and the campaign costs about 54 cents a connected call, roughly 22 cents a connected minute before carrier pass-through, numbers and messages, against a headline rate of 12 cents.

The levers, in order of effect: list quality, call length, escalation rate, then the tier.

Is Bland AI any good for enterprise call volumes?

For high-volume regulated calling it is a serious option, provided you buy the Enterprise tier, because concurrency, dedicated infrastructure and the compliance paperwork all live there.

Self-serve plans top out at 100 concurrent lines, while a contact center running collections or claims intake needs more lines plus the contract controls. Teams buying a custom voice agent for a call center buy two things anyway, the platform and the engineers who wire it to the dialer and the CRM. For each workload shape, see our guide to voice AI agent platforms for inbound and outbound calls.

What are the alternatives to Bland AI?

The field splits into bundled-rate platforms, component platforms that pass model and voice costs through, and self-hosted stacks you run yourself. Bland AI alternatives fall into those groups, and the fit depends on how much of the stack you want to own.

Vapi bills usage first: Vapi's pricing page lists $0.05 a minute for hosting, with models and voices passed through at cost. Success packages run $29 a month, or 10% of the hosting fee with a $999 monthly minimum. Choose Vapi to pick each model and voice yourself.
Retell AI publishes every component: Retell AI's pricing page quotes $0.07 to $0.31 a minute, built from $0.055 voice infrastructure, voices from $0.015, your model, and $0.015 telephony. Choose Retell to tune each line of the stack.
Synthflow publishes only an enterprise tier: Synthflow's pricing page states that enterprise contracts start at $30,000 annually, scoped on volume, concurrency, telephony and security review. Choose Synthflow for launch support inside one contract.
A self-hosted stack, with telephony, speech and the model on infrastructure you own, trades a usage rate for capacity you buy once. Security teams reach for it when call audio cannot leave the network.

How do you compare voice AI quotes like for like?

Normalize every quote to one number, your cost per connected minute at your own answer rate, before comparing anything else.

Define the workload: monthly attempts, answer rate, average connected minutes, transfer rate, messages per call.
Ask what the minute includes. Bland voice AI bundles the model, transcription and speech output; component platforms charge each separately.
Price telephony on its own. Bland's per-attempt minimum applies on its own telephony; on your own carrier, price that carrier's rates instead.
Add the fixed fees: platform or success packages, numbers, extra concurrency, compliance add-ons.
Check the caps and gated features, since concurrency, daily limits, SSO, a BAA and data residency decide whether a plan fits.
Divide by connected calls. The winner is the lowest cost per outcome, not the lowest rate.

What mistakes should you avoid when budgeting Bland AI pricing?

Four mistakes account for most overruns:

Budgeting the headline rate. The example above lands near double it.
Ignoring unanswered attempts. A poor list spends the minimum charge thousands of times over.
Pricing production on the pilot plan. Concurrency and gated features change the quote.
Forgetting follow-up channels. A confirmation text after every call adds a per-message charge.

How Origins AI runs outbound voice on owned infrastructure

Origins AI (originshq.com) builds Origins AI Voice AI, an inbound and outbound voice agent stack deployed on the customer's own infrastructure rather than consumed as a hosted service. According to its Voice AI product page, it covers telephony over SIP trunking or PSTN, a speech layer with real-time transcription and neural speech output, and an action engine for CRM lookups and escalation. Deployment modes are on-premise, private cloud, hybrid and air-gapped.

The company says latency runs under 300 ms at the 99th percentile in optimized deployments, with up to 500 concurrent sessions a node. It reports a pilot on one use case typically live within 30 days, with its implementation team handling deployment and integration. In on-premise and air-gapped deployments, no call audio or transcript leaves your network; in hybrid mode the context sent to a hosted model does. Origins AI does not publish a rate card, so the comparison is a usage rate against capacity you own.

Our guide to B2B versus B2C AI cold calling covers consent and disclosure, and the Voice AI versus Vapi comparison sets a self-hosted stack against a hosted platform.

Talk to an engineer

Send one campaign brief, with attempts, answer rate, call length and escalation rate, and we will scope what it takes to run it on infrastructure you own. Book a call with an engineer.