July 17, 2026. The free ride on Meta's AI agent for WhatsApp has an end date, and it is two weeks away. From August 1, businesses using the Meta Business Agent on WhatsApp will be billed by token usage rather than per message, at a global rate of $2.00 per million tokens, as Business Today reported from Meta's announcement earlier this month. The agent itself launched at Meta's Conversations event on June 3 and answers customer questions, recommends products, books appointments, captures and qualifies leads, and hands off to a human when needed, across WhatsApp first.
The key numbers
- $2.00 per million tokens from August 1. Meta suggests a typical message can consume up to 25,000 tokens, which works out to roughly 4 to 5 cents per AI-handled message, with final cost depending on interaction complexity. The token charge bundles AI processing and message delivery into one line.
- Smaller businesses get a different door. Access for smaller operations comes through select tiers of the WhatsApp Business Premium subscription, while larger businesses pay by token usage.
- Service message fees return October 1. After nearly two years free, non-template service messages sent outside the Business Agent will again be charged at utility and authentication rates, which vary by market. Meta says it will publish the October 1 rates by September 1 and may revise pricing up to once per quarter.
What it means for operators
This is the same shift we mapped when the whole AI stack went usage-metered: the platform gives you a free window to build a habit, then the meter starts. The difference here is that you have two more weeks of free production data, and that data is exactly what you need. Run the agent hard between now and July 31 and measure three numbers per conversation: tokens consumed, resolution rate without human handoff, and revenue events, bookings or qualified leads. At 4 to 5 cents per message, a WhatsApp agent that books appointments is trivially cheap. One that burns 25,000 tokens to frustrate a customer into asking for a human is not. Know which one you have before August 1, and set the comparison honestly against alternatives: a GoHighLevel WhatsApp automation with fixed platform pricing, or a custom flow where you control the model and the token bill. For teams doing serious volume on WhatsApp, this billing change is worth an afternoon of math now rather than a surprise invoice in September. Our AI automation team runs exactly this kind of channel cost audit, token burn versus booked revenue, across Meta's agent, GHL, and custom builds.
Frequently Asked Questions
Meta will bill $2.00 per million tokens globally, replacing the per-message model. Meta's guidance suggests a typical message can use up to 25,000 tokens, roughly 4 to 5 cents per message, though complex interactions cost more. The charge combines AI processing and message delivery. Smaller businesses access the agent through select WhatsApp Business Premium tiers instead.
Use the remaining free period to gather production data: measure tokens per conversation, how often the agent resolves issues without human handoff, and how many bookings or qualified leads it generates. Those three numbers tell you whether the agent pays for itself at metered rates, and they are only free to collect until July 31, 2026.
Non-template service messages, the free-form replies businesses send inside the customer service window outside the Business Agent, will be charged again after being free for nearly two years. They will be billed at utility and authentication message rates, which differ by market, and Meta says it will publish the applicable rates by September 1, 2026.
It depends on volume and control. Meta's agent is fast to deploy and now has transparent per-token pricing, but you do not control the model or the token consumption. Platforms like GoHighLevel offer WhatsApp automation under fixed platform pricing, and custom builds let you choose cheaper models for high-volume flows. The right answer comes from comparing cost per resolved conversation across all three.