Bland AI Review (2026)

Bland's pricing page lists three things as included in the per-minute rate: the language model, speech to text, and synthesis. Telephony is not among them, and the FAQ underneath says so directly — telephony is billed separately, on your own carrier or Bland's at pass-through cost.

The documentation site publishes a machine-readable index for AI agents to read. It describes the pricing page as offering per-minute pricing that bundles the model, synthesis, speech to text and telephony, with no per-token charges.

Those are the same company describing the same page, and they do not agree. We read both on 2026-08-06.

The $0.09 that competitors still quote

Bland is widely cited at $0.09 a minute. That was a real, purchasable, flat rate, and it was retired on 5 December 2025 when the vendor moved to plan-based pricing.

It does not appear on the pricing page today. Where it survives is in comparison articles, most of them published by competitors and affiliates who have not re-checked since. An earlier version of this review implied Bland was advertising a rate nobody could buy, and that was unfair: the vendor dated the change, published the new rates, and published the arithmetic of what it would cost people.

That arithmetic is worth reproducing, because a vendor rarely publishes it. Bland's own transition note works through a customer spending $40 a month at the old rate, which is 444 minutes, and shows the same usage costing $62.22 at the new Start rate. A 56% increase, printed by the vendor, with one month of the difference refunded as a credit.

What the platform fees actually change

Three self-serve tiers, and the fee is the thing to model rather than the rate.

The headline rate on this site is Scale's $0.11, and it is the cheapest per minute. It is not the cheapest bill until you are running a lot of minutes.

Read the top row twice. Scale is the plan carrying the lowest advertised rate, and below 20,000 connected minutes a month it costs more than the tier beneath it. Twenty thousand minutes is 333 hours of live conversation. Most teams evaluating this product are an order of magnitude below that, which means the rate this site quotes, and that every comparison quotes, is the wrong rate for most of the people reading it.

At a more typical 5,000 minutes a month, Scale's $499 fee plus $0.11 a minute works out at $0.21 a minute effective, before telephony. The platform fee nearly doubles the meter.

The vendor's own example bills twelve minutes for a ten-minute call

Bland publishes a worked example: a ten-minute call where the agent talks for eight minutes and then transfers to a human for two. On Scale it charges ten minutes of connected time at $0.11 and two minutes of transfer time at $0.03, totalling $1.16.

The two transferred minutes are billed twice, once in the connected total and once as transfer. That is the vendor's own table, and it puts the effective rate at $0.116 a minute, five percent above the headline, on a call pattern the product is explicitly built for. Customers who bring their own Twilio pay no transfer fee at all, which makes bringing your own carrier a rate decision rather than a plumbing one.

The architecture is the reason to consider it

Bland runs its own stack end to end rather than orchestrating other vendors' services. Latency tuning happens in one place rather than across four seams, and the invoice is one number rather than five. Against that, you cannot swap a component when something better ships, and you are exposed to one vendor's roadmap at every layer at once.

Two smaller things the pricing page does well and most competitors do not. Billing is prorated to the exact second with no rounding up, where Ultravox rounds to six-second blocks and several vendors round to the minute. And the free Start plan is a plan rather than a trial: no card, two credits, an inbound number, and a 99.9% uptime commitment that applies on the free tier as much as on Enterprise.

The honest warning sits in the vendor's own billing documentation, which states that a negative balance means usage outran the credits before the system could stop it. The stop is not guaranteed, so treat the balance as a soft limit rather than a cap.

Self-serve with a genuinely free tier, and the single-vendor stack means one account rather than four. Getting an agent answering is quick.

• Settle the telephony question in writing first. The pricing page and the documentation index disagree about whether it is in the rate, and it is the difference between 1.13× and 1.00×.
• Start on the free tier regardless of what the sales conversation suggests, until you have measured a full month of real minutes.
• Do not buy Scale for the rate. Below 20,000 connected minutes a month it costs more than Build, and below 16,633 it costs more than paying nothing at all on Start.
• Count transfers separately. Transferred minutes are billed twice on Bland telephony, and not at all if you bring your own Twilio.
• Watch the balance rather than trusting it. The vendor states that usage can exceed available credits before the system stops it.

Unusually complete for this category: three self-serve tiers with rates, fees, concurrency and caps all published, plus a worked example and a dated record of the last increase. Two things are missing and one is contradicted.

Missing: any enterprise figure, and the pass-through telephony rate. The pricing page says telephony is billed at pass-through cost and never says what that cost is. The only telephony prices published anywhere are $15 a month per number and a $0.015 minimum per outbound call.

Contradicted: whether telephony is in the rate at all. This site models it as excluded, which is what the pricing page says, giving $0.124 a minute all-in. Get a written answer before budgeting either way.

One number to carry into the room: the platform fee is month-to-month, not a commitment. There is no annual term, no minimum and no contract, so a tier that stops paying for itself can be dropped the following month. That is worth more than it sounds against vendors here selling four-year terms.