AI tokens in Kotia: text volume, context and cost
Why a token is not a word, how model tokens differ from your Kotia balance, and how to estimate text volume and check the cost of your work.
On this page5
A language model token is a fragment into which the system splits text for processing. A token can be a word, part of a word, a character, or another sequence of symbols. There is no exact rule that one word equals one token. The split depends on the model, the language, and the text.
In Kotia, the word "tokens" also refers to the units you pay with from your balance. This is a different quantity. The difference helps explain why a large file may not fit into the context, while a short answer may require significant processing of the source data.
Two meanings that come up in practice
| Term | What it measures | Where it helps |
|---|---|---|
| Model tokens | The amount of information the model receives and generates | When estimating the context window and the length of the work |
| Kotia tokens | The available funds for paying for tasks and services | In your balance, order costs, and operation history |
You don't need to subtract the number of model tokens directly from your Kotia balance. The cost of work is set by the rules of the model or service you choose.
Why a short answer doesn't always mean low usage
Imagine you shared a detailed report and asked for five conclusions. The result fits in one paragraph, but the source material still has to be processed to produce it. The previous messages and instructions in the conversation may also count.
A model's work has an input (the request and materials) and an output (the generated response). Both matter when estimating the volume. For some Kotia services, the cost may be fixed, so the length of the response alone does not determine the price.
How to check your costs in Kotia
- Before you order a service, read the cost or the preliminary estimate if it is shown in the confirmation.
- For a large text, check the context fill indicator next to the input field. It shows volume, not price.
- Run a typical task with the model and settings you need.
- Open profile → "Tokens" → "Balance history".
- Find the operations by time and task type. If a recalculation happened, look at the related charges and refunds together.
- Use the actual usage as a guide for similar work, taking into account differences in the source data and the size of the result.
In workflows, the costs of several steps add up. For details, see workflow costs.
How to send less unnecessary data
If you need a conclusion about one section of a document, specify that section and pass the matching fragment. Remove repeated instructions, unrelated examples, and attachments that are not connected to the task. At the same time, keep the context without which the figures or conditions could be misread.
For example, to compare delivery times, it is more useful to pass the table of terms and the related notes, rather than a full catalog with product images. But removing the note about exceptions can distort the answer.
What to do when you hit different limits
If the model doesn't have enough available space, use how to manage the context window: split the materials or move the task to a new conversation. If your Kotia balance has run out, check token types and top-ups.
Buying a package adds to your balance for paying for work, but it does not increase the context window of the selected model. A shorter prompt is not always better either. Keep important data and conditions, even if they increase the volume.