# Understanding why Circle AI credit costs vary by message

https://help.circle.so/p/administration/circle-ai/understanding-why-circle-ai-credit-costs-vary-by-message

Category: [Administration](https://help.circle.so/p/administration) › [Circle AI](https://help.circle.so/p/administration/circle-ai)

---

_Last updated: August, 2026_

You might notice that some messages to Circle AI use more credits than others. In this article, we’ll explore how token-based pricing works.

> ❗ **Important:** Credit cost isn't based on how many messages you send. It's based on how much work each message asks Circle AI to do.

📚 For more information on Circle AI billing, see [**Understanding your Circle AI billing and credits**](https://help.circle.so/p/administration/circle-ai/understanding-your-circle-ai-billing-and-credits) **.**

## About tokens

A token is a small chunk of text (often a whole word or part of a word), used by AI models to read and process language. When you send a message to Circle AI, your text (input) and the response Circle AI provides (output) are broken down into tokens.

Credits are how Circle measures that token use. The more tokens your message and its response contain, the more credits the message costs.

## Why do some messages cost more than others?

The size of a request determines its token count, which in turn drives its credit cost.

Shorter requests require fewer tokens. Circle AI reads your message, writes a brief answer, and uses a small number of credits.

Larger, multi-step requests contain more tokens. For example, setting up a space group with multiple spaces or creating a course curriculum with populated lessons and quizzes. Circle AI has to do more work and produce a much longer response, so the token count climbs, and the message costs more credits.

As an example, asking Circle AI to “Invite John to my community using the email [member@circle.so](mailto:member@circle.so) and make them the moderator of Space group A” requires fewer credits than asking it to “Create a new course space in space group A and set up the lessons using the following structure [INSERT CURRICULUM]”. While both these messages are straightforward, the AI takes far more steps to set up a course space, configure its settings, and build a curriculum, resulting in a higher token count and therefore uses more credits.