WhatsApp Business API cost calculator

Rates checked 1 November 2026

A WhatsApp assistant produces two bills, from two companies, and most pricing pages show you one of them. Meta charges per delivered message after the first 1,000 service messages per phone number each month; your software charges a subscription on top. This works out both, in dollars, dirhams or riyals.

Why there are two bills, and which one grows

Until 1 October 2026 a business that only answered customers paid Meta nothing. Replies inside the 24-hour window a customer opened were free, and the only billed messages were the templates a business sends to start a conversation. That changed: service messages are now billed, with the first 1,000 per phone number per calendar month free and no rollover of what you do not use.

So the structure is now this, and it is worth seeing the two parts separately because they grow at completely different speeds:

• Meta, per delivered message. Grows linearly with your customers, and is the bill nobody budgeted for. At the UAE rate it is about $31 for a business answering 3,000 customers a month — small. At 50,000 it is $770.
• Your software, per month. Grows in steps, and how it steps is the thing to interrogate. A subscription with per-message overage grows smoothly. One with a hard cap stops answering your customers. One metered on contacts or seats can double on a month when your message count did not move at all.

The number worth noticing on any pricing page is therefore not the monthly figure. It is the overage rate, or the absence of one.

The rate card, in dollars, dirhams and riyals

Effective 1 October 2026. One column serves utility, authentication and service messages because both of our sources give the same figure for all three in every market. Marketing templates are the separate, dearer column. Dirham and riyal figures are conversions of the dollar card at the two currencies' pegs — AED 3.6725 and SAR 3.75 — which is exact arithmetic rather than a floating rate that goes stale, and is the reason this page can carry them at all.

Two things in that table surprise people who serve the Gulf. Saudi Arabia is 32% cheaper than the UAE on the same message, so a business with customers in both cannot use one blended figure honestly. And Egypt is 4.4 times cheaper than the UAE, which means an Egyptian customer base changes the arithmetic more than any software decision on this page does.

Meta may change these only on 1 January, 1 April, 1 July or 1 October. Nothing else moves them, which is why a dated page is worth more here than a live one.

Why this only prices one of the meters

A reply count is the one quantity a business can measure before it buys anything. Most of this market does not bill on it, which means no calculator can honestly turn your reply count into another platform's invoice — and any that claims to is guessing on your behalf and calling it arithmetic.

So this page prices Meta's bill, which applies whatever software you choose, and ours. For everything else, here are the meters in use and what each one would need from you first. Work out which one you are being sold before you compare any two prices.

• Monthly active contacts — needs your contact list size. Bills for how many people were in a conversation, resolved or not. A quiet month with a large list costs full price, so a reply count tells you nothing about the invoice.
• Agent seats — needs how many people open the inbox. Bills per person with access, usually with a conversation allowance on top. Two meters at once, and the second is the one that moves.
• Prepaid credits — needs what a credit buys per message category. A credit is worth a different amount against marketing, utility and service messages, and some credit pools expire. None of that converts from a reply count.
• Resolutions — needs your messages-per-conversation ratio. A resolution is several messages and how many depends on your customers. Turning replies into resolutions would put our guess where your number should be.
• Bundled into a subscription — needs a published allowance, which is not always given. Some subscriptions include AI replies without saying how many. Where no allowance is published there is no arithmetic to do, and a low headline price is not evidence of a low bill.
• Answered messages — needs nothing further, which is why this page can price it. Bills per reply actually sent. This is the meter the calculator above uses, and the one Jabee bills on.

What each of those does at the cap is the other half of the question, and what to ask before you buy covers it: some platforms stop replying until the month resets, others keep answering and bill the excess. The monthly price does not tell you which.

If you were building it yourself

A fair number of people reading a page like this are weighing a subscription against writing the thing. The model bill is the part that is hard to find, so here it is at your volume, and both figures are estimates rather than measurements:

Meta's figure comes from Meta: $2.00 per million tokens since 1 August 2026, and Meta's own statement that a typical message runs 20,000 to 25,000 tokens. Ours comes from counting tokens in our own code, which is why we label it an estimate and not a measurement — roughly 6,500 input and 180 output tokens across the two model calls one answered question takes. Meta's service fee is on top of either, and so is everything that is not the model: the retrieval, the webhook handling, the inbox, and the part that takes the real time, which is noticing when the assistant has answered a customer confidently and wrongly.

How this page was made, and what is wrong with it

What is primary-sourced. The free allowance — 1,000 service messages per business phone number per calendar month, with no rollover — is from Meta's own documentation, read on 1 November 2026. So is the quarterly change window, and Meta's $2 per million tokens for Business Agent.

What is not. Every per-message rate on this page. Meta publishes those only inside downloadable CSV and PDF rate cards sitting behind a JavaScript selector; its documentation page states the rates exist and links the downloads without printing a single figure. We checked that again on 1 November 2026. What we have instead is two independent republications that agree on every market they share. One is published by a Meta Business Partner and states it is built on Meta's October 2026 rate cards; the other is a by-country table from a WhatsApp solution provider, itself cross-checked against three further vendor tables in our own research. They match exactly on the UAE, Saudi Arabia, Egypt, India, North America and the UK. We have stopped naming other companies on this site, so the two are described rather than linked — if you want to audit the figures, the thing to audit them against is Meta's own downloadable card, not somebody else's copy of it.

Where they disagree, and we have not picked a winner. Qatar and Kuwait. One source gives them dedicated rows — Qatar at $0.0120 service, Kuwait at $0.0440 — and the other says both fall into the Rest of Middle East grouping at $0.0091. That is a four-fold difference on Kuwait and we cannot resolve it, so both countries are served here as Rest of Middle East and this paragraph exists instead of a confident number. If you are in either, use the override field with the rate on your own Meta invoice.

Three more honest weaknesses. The dirham and riyal columns are conversions of the dollar card at the currencies' pegs, not Meta's own AED and SAR cards, which we could not download. Competitors' subscription figures are their published terms and not our experience of paying them, because we have run a paid account on almost none of them. And we sell one of the platforms in the table, which no disclosure fully fixes — the useful test is whether the tool ever tells you to buy somebody else, and at 1,000 replies and below, which is where most businesses reading this actually are, it does.