WhatsApp Business API Pricing in 2026: Every Rate, and Exactly What Changes on 1 October

Direct answer. Meta charges per delivered message, priced by the message category and the recipient's market. On 1 October 2026 two things that are free today start costing money: free-form replies you send inside the 24-hour customer service window, billable after the first 1,000 delivered per phone number per calendar month, and utility templates delivered inside that window, billable from the first one. The 24-hour window itself does not change. Rates checked on 17 September 2026 against Meta's published rate cards.

How WhatsApp Business API pricing works in 2026

Since 1 July 2025 the WhatsApp Business Platform has billed per message rather than per 24-hour conversation. Four things determine what you pay.

• The category. Marketing, utility, authentication, and service. Marketing is always the most expensive; service and utility sit at the bottom.
• The recipient's market. Not where your company is registered. An Italian business messaging a number in the United Arab Emirates pays the UAE rate.
• Delivery. You are billed on delivered messages, not on API calls. A message that never lands is not a billable message, which is why your cost accounting should read WhatsApp delivery events rather than counting POST /messages requests.
• Your monthly volume. Utility and authentication messages fall into cheaper tiers as you send more within a calendar month. Marketing has no volume tiers, and service messages, as of October, have none either.

Meta publishes four rate cards: marketing, utility, authentication, and authentication-international. There is no separate service card, because Meta defines the service rate as identical to the utility and authentication rate for that market. You can pull the raw cards per currency from Meta's pricing documentation or browse them interactively on the WhatsApp Business pricing page.

What changes on 1 October 2026

Meta announced the change on 10 August 2026 and published the new rates on 1 September. It takes effect on 1 October 2026 at midnight in each WhatsApp Business Account's own timezone, so a Sydney account crosses over before a Los Angeles one.

Read past the headlines on this one. A lot of coverage says the 24-hour free window is being removed. It is not. Meta's own wording is that the customer service window "opens and resets with each user message", exactly as it does today. What changes is the billing inside it. The distinction matters because your permission model stays the same while your cost model does not: 24 hours still means you may send free-form messages, it just no longer means those messages are free. Zendesk's announcement to its customers frames it the same way.

The 1,000 free service messages, explained properly

This is the part most write-ups either miss or get wrong, and it is the one that decides your invoice.

• It is 1,000 messages, not 1,000 conversations and not 1,000 customers.
• It is per business phone number. Run four numbers and you have 4,000 free service messages a month.
• It counts delivered messages, and charging starts at the 1,001st.
• It resets on the calendar month and unused messages do not roll over.
• It covers service messages only. Utility templates get no allowance at all.

Here is why message fragmentation suddenly has a price. A customer asks where their order is. Your agent, or your bot, replies like a human being:

1. "Hi John"
2. "I checked your order."
3. "It shipped yesterday."
4. "UPS should have it with you tomorrow."
5. "Here is your tracking link."

That is five delivered service messages, not one conversation. Before October it costs nothing. After October, once that number has used its monthly thousand, it is five billable messages for one answer. The same answer written as one well-composed message is one billable message. Nothing about your product changed; your bill just dropped by eighty percent on that exchange.

Meta’s official pricing documentation confirms 1,000 free delivered service messages per business phone number per calendar month, shared across recipient markets, with no rollover. The same allowance is described by 360dialog and respond.io. Check your provider’s invoice terms separately: platform fees and any provider markup are not Meta’s service-message charge.

What a service message costs in your market

Service messages are priced at the same list rate as utility and authentication messages for the destination market. The critical asymmetry is that service messages get no volume tiers. Meta's documentation states it plainly: there are no volume tiers for service messages. Send a hundred or a hundred thousand, the per-message rate is flat. Utility and authentication keep their tiered discounts.

Two things worth absorbing from that table. Germany costs thirty-nine times what India costs for the identical message, so a support team serving Mumbai and one serving Munich are running completely different businesses after October. And Europe is expensive: most of Western Europe sits between two and five and a half cents per reply, which is where a fragmented inbox becomes a real line item rather than a rounding error.

Utility templates inside the 24-hour window

This change is quieter than the service one and easier to miss in your own billing.

Since 1 July 2025, a utility template delivered while the customer's 24-hour window was already open has been delivered without Meta's utility charge. That exemption ends on 1 October 2026. So this exchange:

• Customer: "Where is my package?"
• You: an approved order_status utility template

is free of Meta's template fee today, and billable from 1 October even though the customer started the conversation and the window was wide open. Unlike service messages, there is no free allowance to absorb the first thousand. Utility templates are chargeable from the first delivered message. The compensation is that utility keeps its volume tiers, so high senders still see the rate fall as monthly volume climbs.

Practical consequence: audit which of your templates fire inside open windows. Many teams send a utility template for something a plain text reply would handle, because until now both were free. From October one of those is tier-discounted and the other eats your service allowance, and which is cheaper depends entirely on your volume.

Two clocks: the 24-hour window and the 72-hour free entry point

These get conflated constantly, and conflating them will cost you money or get your message rejected.

The 24-hour customer service window is a permission. A customer messages you, it opens, and every new message from them resets it. While it is open you may send free-form, non-template messages. When it closes you need an approved template to make contact again. From October, the messages you send while it is open are service messages and are billable past your allowance.

The 72-hour free entry point is a billing exemption. When someone reaches you from a Click to WhatsApp ad or an eligible Facebook Page call-to-action button and you reply within 24 hours, a 72-hour window opens in which message delivery is free across every category, service included. Meta's documentation states 72 hours, and every provider we checked agrees. If you see a claim that this is becoming seven days, treat it as unconfirmed: it is not in Meta's live documentation, and the vendor pages it was attributed to still say 72 hours.

Now the trap. Suppose a free entry point is active but the customer has not written to you for more than a day. The 72-hour exemption can still be running while the 24-hour service window has already shut. You cannot simply type a free-form reply, because the permission has expired. You need an approved template, and that template's delivery charge is waived while the entry point lasts. Two clocks, two jobs: one governs what you are allowed to send, the other governs what you pay for it.

Meta Business Agent versus your own AI agent

Since 1 August 2026, replies generated by Meta's own Meta Business Agent are priced on tokens rather than per message: one global rate of $2.00 per million tokens, worldwide, which works out to roughly four to five cents for a typical reply. Meta applies a single charge covering both the AI processing and the message delivery, so a Business Agent reply is never also billed as a service message.

Everything else is a service message. A human agent typing, your own chatbot, a third-party AI platform, or an assistant you built yourself: from 1 October those replies are service messages under the normal rules and the normal allowance.

The comparison is less obvious than it looks. At $0.0014 in India, it takes about thirty service messages to match the cost of one Business Agent reply. At $0.0550 in Germany, a single service message costs more than a Business Agent reply. And one detail catches people out: Business Agent token charges are not waived inside the 72-hour free entry point. Delivery is free there, the AI processing is not.

Which countries change on 1 October

Alongside the service change, Meta is reshaping the rate card itself. Nine markets leave their regional groupings and become standalone, each also picking up an authentication-international rate above its current regional authentication rate.

Marketing rates rise in Kuwait, Mexico, Morocco, Saudi Arabia, the United Arab Emirates, Rest of Middle East and Rest of Asia Pacific. Utility and authentication rates rise in Kazakhstan, Kuwait, Morocco, Oman, Pakistan, Peru, South Africa and Ukraine, and fall in Bangladesh, Iraq, Nepal and Sri Lanka.

Note that those are two different lists. Pakistan, Peru and South Africa see utility and authentication rises without becoming standalone markets. Mexico, Saudi Arabia and the UAE see marketing rises without becoming standalone. And because the service rate follows the utility rate, every market in the utility column above is also a service-cost change for you from the same date.

If you send authentication traffic across borders, check the authentication-international column rather than assuming the standard authentication rate. Sri Lanka's international authentication rate is seventy-two times its domestic one.

What your provider adds on top

Meta's rate card is Meta's fee. It is not your invoice.

Whether you go through Twilio, Sinch, 360dialog, Wati, Freshworks, or a CRM with WhatsApp bolted on, your provider may add a per-message markup, a monthly platform fee, per-seat charges, a monthly-active-user fee, separate automation or AI charges, and tax. Twilio's own WhatsApp pricing, for instance, applies independently of Meta's new October charges.

Here is a concrete way to catch it. While researching this article we compared several published October rate tables against Meta's list rates. One reseller's table sits at 1.10 times Meta's list rate across every market we compared: Germany at $0.0605 against Meta's $0.0550, Italy at $0.0330 against $0.0300, the Netherlands at $0.0550 against $0.0500. That is a clean ten percent markup, disclosed nowhere on the page. Another provider's table ran higher still with no consistent ratio at all. Neither is dishonest, they are retail prices, but neither is Meta's rate either.

So when you are handed a number like $0.0034, ask one question: is this Meta's list rate, or is this what you will invoice me? Then ask for the October price list in writing before the first of the month.

Rather than work this out by hand, put your own numbers into the WhatsApp pricing calculator: it prices your markets on the same rate card, in three steps.

A worked example: one support inbox, before and after

Assumptions, all hypothetical and chosen to be easy to re-run with your own figures. One UK business phone number. 2,000 inbound customer conversations a month. An average of 3.5 free-form replies per conversation, which is typical of a chatty support inbox. 600 order-status utility templates sent inside open windows. UK service and utility rate: $0.0220.

Three readings of that table. First, a line that was genuinely zero becomes a real monthly cost, and it scales linearly with how talkative your inbox is. Second, cutting the average from 3.5 replies to 1.8 by writing complete answers halves the bill, $74.80 a month saved on one number, without reducing service quality or message volume in any way that a customer would notice. Third, and easy to overlook, none of this is affected by volume discounts, because service messages do not have any. Growing your way into a cheaper rate is not available here. The only levers are fewer messages and better routing.

Multiply by however many numbers you run. Ten numbers with this profile is $1,452 a month appearing on a bill that read zero in September, or $704 if your replies are tight. If you want the same model run against your own conversation volume and reply count, book a pricing walkthrough and we will build it with you.

What to do before 30 September 2026

1. Attach a payment method to every WhatsApp Business Account and phone number. This is the one with a hard consequence. Providers relaying Meta's instructions state that any Solution Provider or directly-integrated business without a payment method on file by 30 September 2026 will have Meta stop delivering service messages once they become chargeable on 1 October. Check your Billing Hub, check your provider's credit arrangement, and check who receives the billing failure alerts.
2. Decide what your application does when a service message is rejected for billing reasons. If you have no handling for it today, you will find out in production.
3. Consolidate your replies. One complete message instead of five fragments. This applies to human agents, canned replies and any AI in the loop. It is the largest single saving available and it costs nothing to implement.
4. Audit templates that fire inside open windows. Anything that was free because the window was open is now a cost centre. Decide per template whether a tier-discounted utility template or an allowance-consuming free-form reply is cheaper at your volume.
5. Capture the free entry point. If you run Click to WhatsApp ads, make sure you reply within 24 hours so the 72-hour exemption actually activates. That is free delivery across every category, and it is the only remaining genuinely free window at scale.
6. Get your provider's October price list in writing, with its markup stated separately from Meta's rate.
7. Move your cost accounting to delivery events. Billing follows delivered messages, so counting API requests will overstate your spend and hide failures.

Where MessageMind fits

MessageMind runs on the official WhatsApp Business Cloud API hosted by Meta, so the rates above are the rates you pay, and it puts WhatsApp in the same inbox as voice, SMS, email, Instagram, Messenger, web chat, Viber, TikTok and PMS inboxes. The AI answers in real time by default, or on a delay you set, handles booking inside the thread with tappable slots, detects and replies in 90 or more languages, and we write and submit your templates for Meta approval so deliverability stays visible in your dashboard.

On cost, the honest framing is that no platform can change Meta's rate card. What a platform can change is how many messages it takes to answer a question, and whether an update goes out as a template or as a reply. If you want to see what that looks like against your own numbers, the WhatsApp use case walks through the flows and our pricing shows what sits on top of Meta's fee. For the neighbouring reading, WhatsApp Business AI automation covers the buying decision and AI customer service ROI covers the maths on the other side of the ledger.

Frequently asked questions

Does the 24-hour window go away on 1 October 2026?

No. It opens when a customer messages you and resets with each new message from them, exactly as it does now, and it still governs whether you may send free-form messages. What changes is that the messages you send inside it become billable past the free allowance.

Are inbound messages from customers still free?

Yes. Messages a customer sends to your business have never been charged and are not being charged from October. You pay for what you deliver, not for what you receive.

What counts as a service message?

Any non-template, free-form message you deliver to a customer inside the open 24-hour window, whoever or whatever composed it: a human agent, your own chatbot, or a third-party AI. Replies from Meta's own Business Agent are the exception and are billed on tokens instead.

Do volume discounts apply to service messages?

No. Meta's documentation states there are no volume tiers for service messages, so the rate stays flat at any volume. Utility and authentication keep their tiered pricing.

Is there really a free tier of 1,000 service messages?

Meta confirms it in its official pricing documentation: 1,000 delivered service messages per business phone number per calendar month, shared across recipient markets. The market rate applies from the 1,001st message and unused messages do not roll over. Utility templates and provider platform fees are separate.

What happens if I have no payment method on file?

Providers relaying Meta's guidance say that a Solution Provider or directly-integrated business without a payment method by 30 September 2026 will have Meta stop delivering chargeable service messages from 1 October. Inbound and other categories are governed by your normal billing arrangement, but this is not a warning to leave until the last week.

Does the 72-hour Click to WhatsApp ad window change?

No. It remains 72 hours of free message delivery across all categories, activated by replying within 24 hours to someone who reached you from an eligible ad or Page button. Claims that it is becoming seven days are not supported by Meta's live documentation.

How are AI agent replies charged?

Meta's Business Agent is charged at $2.00 per million tokens globally, a single charge covering processing and delivery, and is never also billed as a service message. Any other AI agent's replies are ordinary service messages at your market's rate.

Is this the same as WhatsApp Business app pricing?

No. All of this concerns the WhatsApp Business Platform, the API you reach through a provider. The free WhatsApp Business app on a phone is unaffected by the October change.

If you are sizing this for October right now, the two numbers you need are your monthly delivered service messages per phone number and your average replies per conversation. Bring them and book a pricing walkthrough, and we will run your markets against the rate card and show you what consolidation is worth on your volume.