The official WhatsApp API is no longer text only. Since Meta released WhatsApp Business Calling, your company can call the customer inside WhatsApp itself, over voice on the internet. And that has its own price, separate from the price of messages.
Good news for anyone worried about the bill: an inbound call from the customer is free. You only pay when the business places the call. In Brazil, the rate starts at R$0.0556 per minute, billed in 6-second pulses.
In this guide you will see what gets charged, the official table by volume tier, the effect of 6-second pulses on your invoice and a monthly cost simulation for a mid-sized Brazilian operation.
Quick summary
- A call started by the customer is free. Only a call started by the business is billed (Meta calling pricing documentation, consulted in August 2026)
- In Brazil, the entry rate is R$0.0556 per minute of business-initiated calling, up to 50,000 minutes a month (Meta’s official BRL rate card, effective since July 1, 2026)
- Billing is by duration, in 6-second pulses, and a partial pulse counts as a full one: a 56-second call is billed as 60
- Calling cost adds to messaging cost. The permission request is free inside the customer service window and becomes a paid template outside it
- You can request permission once every 24 hours for the same customer, up to twice a week
- Calling is cheap next to broadcasting: 750 minutes of calls cost less than 130 marketing sends
- You can put AI on the call. With realtime voice models the conversation sounds natural, and inbound calls still cost R$0 in Meta rates
how much the official WhatsApp API costs: full pricing guide
How Meta charges for WhatsApp calls
Call billing has nothing to do with message billing. They are two separate accounts on the same invoice.
Three rules define the amount:
- Who started it. A call the customer places to the business is free. A call the business places to the customer is billed.
- How long it lasted. The price is per minute, measured in 6-second pulses.
- Which country. The rate follows the country code of the number being called, not of your own number.
Add the volume tier to that: the more minutes your company accumulates in a calendar month, the lower the per-minute rate. We cover that in the table below.
One technical detail that matters: WhatsApp Business Calling only works through the Cloud API, the version of the official API hosted by Meta. It does not work in the regular WhatsApp Business app. If you use an automation platform connected to the official API, the feature is already on the right track.
It is also worth knowing that business-initiated calling is not enabled everywhere. Meta lists the United States, Canada, Egypt, Vietnam and Nigeria as exceptions (Meta calling documentation, consulted in August 2026). Brazil is enabled.
What is free and what is billed
This is the part that causes the most confusion when you build a budget:
- Customer calls the business: R$0. No rate, whatever the duration. If your operation is inbound service, your voice cost is zero.
- Business calls the customer: billed per minute. It is the only voice item that shows up on the invoice.
- Permission request to call: billed as a message. Before calling someone, your company has to ask for authorization. That request travels as a message and falls under the per-message price table, not the voice one.
In other words: the cost of an outbound call is always permission request + minutes spoken. Budgeting the minutes alone underestimates the bill.
There are frequency limits worth knowing, and they apply per customer, not per number. For the same customer, the business can send one permission request every 24 hours, up to two in seven days. The cap on business-initiated calls, meanwhile, went from 10 to 100 per day in a December 2025 update (Meta calling documentation, consulted in August 2026).
In practice, the limit does not block volume: it blocks nagging. You can talk to as many people as you want, you just cannot ask the same person for permission twice in one day. For small and medium businesses, the cap rarely shows up.
Call price table for Brazil (August 2026)
The figures below come from Meta’s official BRL rate card, effective since July 1, 2026. Meta prices calling per country, so this table applies to calls placed to Brazilian numbers; for another destination country, check that country’s rate card. The tier is set by the total business-initiated calling minutes accumulated in the calendar month.
| Minutes in the month (outbound calls) | Rate type | Price per minute (Brazil) |
|---|---|---|
| Up to 50,000 | List rate | R$0.0556 |
| 50,001 to 250,000 | Tier rate | R$0.0350 |
| 250,001 to 1,000,000 | Tier rate | R$0.0227 |
| 1,000,001 to 2,500,000 | Tier rate | R$0.0175 |
| 2,500,001 to 5,000,000 | Tier rate | R$0.0103 |
| Above 5,000,000 | Tier rate | R$0.0093 |
Source: Meta’s official BRL rate card for calling, effective since July 1, 2026. How to download it: Meta’s calling pricing page → rate cards section → BRL file. The figures change without notice; confirm at the source before signing off on a budget.
Read the table honestly about your own size. The discounted tiers look generous, but the first one only starts at 50,000 minutes a month. That is more than 833 hours of outbound calling, every month. No small or medium business comes close. In practice, for almost every Brazilian business the real price of calling is a single number: R$0.0556 per minute.
One rule in favor of companies that grow: when a call crosses the boundary between two tiers, Meta bills the entire call at the lower rate.
The effect of 6-second pulses
Meta does not charge per second or per full minute. It charges in 6-second pulses, and a pulse started is a pulse billed.
The example from Meta’s own documentation: a 56-second call equals 9.33 pulses and is billed as 10 pulses, that is, 60 seconds.
Each pulse costs R$0.00556 in Brazil. It looks irrelevant, and it is, when the call is long. But the rounding weighs on short calls, because the waste is proportionally bigger.
Compare two operating patterns with the same number of calls:
- 100 calls of 8 seconds (lightning confirmation, the customer picks up and hangs up): each one turns into 2 pulses, that is, 12 seconds billed. You talk for 800 seconds and pay for 1,200. Fifty percent more.
- 100 calls of 5 minutes (real conversations): 300 seconds land on exactly 50 pulses. Zero waste.
The practical lesson is not to avoid short calls, it is to stop using voice for something a message handles better. Confirming an appointment is a utility template at R$0.0350. Calling for that costs more and annoys more.
Practical simulation: how much per month
Let’s simulate a clinic that uses voice for what voice is good at: reconnecting with people who went quiet in chat. The rates below are the Brazilian ones, so treat this as a Brazil scenario.
Scenario: 300 outbound calls in the month, averaging 2 minutes and 30 seconds.
- Minutes spoken: 2 min 30 s is exactly 25 pulses, with no rounding. 300 calls × 2.5 min = 750 minutes.
- Cost of the minutes: 750 × R$0.0556 = R$41.70.
- Permission requests: 300 messages, billed under the message table. Here the amount depends on the customer service window. With the window open, the request goes out as a service message and today costs nothing. With the window closed, it becomes a template and costs R$0.0350 in the utility category, or R$10.50 in total.
Month total in Meta rates: between R$41.70 and R$52.20, depending on whether the requests go out inside or outside the service window. On top of that sits your automation platform plan, which is the same with or without calling.
Mark a date: on October 1, 2026, service messages start being billed at the same utility rate. From then on, both paths cost the same R$10.50, and the window stops discounting the permission request.
how the 24h and 72h WhatsApp windows work
Now the comparison that gives you the real scale: those R$41.70 of voice are worth fewer than 130 marketing sends, which cost R$0.3217 each in Brazil (Meta’s official BRL rate card, effective since July 1, 2026). Three hundred calls of two and a half minutes come out cheaper than one small campaign.
And to reach the second pricing tier, this clinic would need to multiply its volume by almost 67.
marketing vs utility messages: how to classify and pay less
When calling pays off and when it does not
Voice is cheap per minute but expensive in attention: yours and the customer’s. Four criteria to decide:
- Use voice where writing stalls. Negotiation, objections, complex scheduling, delicate collections. Where the conversation goes back and forth, two minutes of talking beat twenty messages.
- Do not use voice for notices. Confirmations, codes, order status and reminders are utility templates at R$0.0350. Cheaper, less invasive, and the customer reads them whenever they want.
- Prefer that the customer calls you. Inbound calls are always free. Leaving a visible call button in the flow drops the cost to zero and gets you a customer who is already interested.
- Ask for permission with context. The permission request costs money and burns patience. Sending requests to a cold list spends twice: the message, and the chance that the customer flags you as a nuisance.
Automation helps exactly here: qualify in chat, handle the repetitive part in text and escalate to voice only when the conversation shows it is worth it.
Can you put an AI on the call?
You can, and it moves the math more than you would expect.
Meta does not care who is on the line. The rate is per minute of business-initiated calling, whether a human or an AI is doing the talking. What the AI changes is what becomes viable.
Start with the obvious side: inbound calls are still free. An AI that picks up the phone absorbs demand spikes, nights and weekends with zero voice cost on the Meta invoice. Where the choice used to be between letting it ring and paying for after-hours staff, there is now a third path.
What unlocked this was the generation of realtime voice models, the so-called speech-to-speech approach. Before, AI on the phone was a three-part chain: transcribe the audio, think in text, synthesize the reply. Every stage added delay, and you got that odd silence before each sentence. Realtime models, such as the OpenAI Realtime API, listen and speak directly, with no text step in the middle. You can interrupt the AI mid-sentence, the way you would with a person. That is the difference between sounding like a phone tree and sounding like support.
In NicoChat, voice is a channel type, the same way WhatsApp and Instagram are. The call lands in the same flow, the same contact and the same history as the rest of your service: the AI answering the phone sees what the customer already said in chat, applies the same tags, moves the card on the Kanban and hands off to a human agent when the conversation calls for it. That is 18 channel types in one operation, and you can try it free for 7 days, no credit card, at nicochat.com.
One budget warning, since this is a post about cost: the voice model is billed separately, by the AI provider, per minute of audio processed. It does not show up on the Meta invoice or in your platform plan. So an outbound AI call has three lines, not two: permission + Meta minutes + model minutes. Confirm the provider’s current rate before projecting volume, because model pricing moves fast.
And the criterion from the previous section does not change with AI. If a message solves it, the message is cheaper. AI on the phone is for conversation, not for notices.
Frequently asked questions about WhatsApp calling costs
Are WhatsApp Business calls free?
It depends who calls. A call the customer places to the business is free, with no duration limit. A call the business places to the customer is billed per minute: R$0.0556 in Brazil, in the first volume tier.
Do I need permission to call the customer?
Yes. The business has to request authorization before making an outbound call, and that request is a message billed under the normal message table. Meta also limits how many requests you can send per day, which rules out mass voice campaigns.
Does calling work in the WhatsApp Business app?
No. WhatsApp Business Calling requires the Cloud API, the version of the official API hosted by Meta. Anyone using only the regular app does not have the feature.
How does Meta measure call duration?
In 6-second pulses, always rounded up. A 56-second call counts as 60. That makes very short calls proportionally more expensive than long ones.
Is it worth swapping marketing sends for calls?
On pure cost, calling is cheap: 750 minutes cost less than 130 marketing sends. But the channels serve different purposes. Calling works for conversations that need negotiation; broadcasting works for reach. Swapping one for the other without looking at intent usually makes both results worse.
How much extra does it cost to put AI on the call?
Meta’s rate does not change: it is the same per minute with a human or an AI on the line. What gets added is the voice model cost, billed by the AI provider per minute of audio, outside the Meta invoice. On inbound calls, which are free in Meta rates, the model becomes the only variable voice cost.
What to take from this read
- Inbound calls are free; only outbound calls from the business are billed
- In Brazil, the math is R$0.0556 per minute for practically any small or medium business, because the second tier only starts at 50,000 monthly minutes
- Billing is in 6-second pulses rounded up, which makes short calls proportionally more expensive
- The real cost of an outbound call is permission + minutes, never just the minutes
- Voice is cheap; use it where text stalls, and leave the repetitive part to utility templates
- With realtime voice AI, free inbound calling turns into real service capacity, but it adds a third line to the budget: the model’s minutes
To build the complete official API budget, with message rates, platform costs and a monthly simulation, see this post’s sibling guide.
how much the official WhatsApp API costs: rates, extra costs and a simulation
