> For the complete documentation index, see [llms.txt](https://help.tellius.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.tellius.com/tellius-6.4/settings/kaiya-and-ai/kaiya-cost.md).
# Kaiya Cost
See what your workspace's Kaiya usage costs, set the per-million-token rates for each model, and cap monthly spend for the workspace and for each user.
Kaiya Cost shows what your workspace's Kaiya usage costs. Each time Kaiya calls a language model, Tellius records which model it used and how many tokens the call took. When you open the report, Tellius works out the cost from each model's rates.
Use this page to see where Kaiya spend comes from, to keep the rates right for the models your workspace uses, and to set monthly spending caps. Find it under **Settings → Intelligence → Kaiya & AI → Kaiya Cost**. The page has three tabs: **Monthly Usage**, **Model Pricing** and **Limits**.
{% hint style="info" %}
**Replaces Kaiya Credits.** In Tellius 6.4, Kaiya Cost replaces the **Kaiya Credits** and **My Kaiya Credits** pages. Usage is shown as cost in US dollars instead of credits, and caps are set in US dollars. Links to the old credits pages in Tellius open Kaiya Cost.
{% endhint %}
### **Who can see each tab**
* **Monthly Usage:** every user can open it. It shows usage for the whole workspace.
* **Model Pricing** and **Limits:** shown to Admins and Super Users. On **Model Pricing**, only Super Users can add, change or delete rates. Admins can view them.
### **How cost is worked out**
For every Kaiya call, Tellius records the model and four kinds of token count:
| Token count | What it covers |
| ------------ | ----------------------------------------------------------- |
| Input | Tokens sent to the model that were not read from the cache. |
| Cached input | Input tokens the model read from its cache. |
| Cache write | Input tokens written to the cache. |
| Output | Tokens the model sent back, including any reasoning tokens. |
Tellius multiplies each count by that model's rate per million tokens and adds the results. All costs are in US dollars.
### **Read the Monthly Usage tab**
**Choose the period.** Select **Month to date**, **Last 30 days** or **Last 90 days** at the top right. Month to date starts on the first day of the current calendar month. The other two count back from today. Every figure on the tab follows the period you choose. Select **Refresh** to load the latest usage without leaving the page.
**Read the two totals.**
* **Total cost** is what all Kaiya usage in your workspace cost over the period, in US dollars.
* **Total tokens** adds up every token Kaiya used in the period, across all four kinds of token count, in millions. A model's tokens count here even if the model has no rates yet.
**Read Cost by model.** The table has one row for each model Kaiya used in the period:
* **Model** is the model's name, as Kaiya called it.
* **Provider** is the service Kaiya reached the model through, such as `openai`, `anthropic` or `openrouter`.
* **Total tokens** is that model's tokens for the period, in millions.
* **Cost** is that model's cost for the period, in US dollars.
Select a column heading to sort the table by that column, for example to put the most expensive model first.
An **unpriced** badge next to a model means it has no rates on the **Model Pricing** tab yet. Add its rates there so all of its usage is costed. See [Set model rates](#set-model-rates-on-the-model-pricing-tab).
#### **Models used through another provider**
Some models are reached through another provider, such as Snowflake Cortex, Bedrock, Vertex AI or Azure OpenAI. A model used this way appears in its own row, separate from the same model used directly. For example, the same Claude model can appear once with the provider `anthropic` and once with `anthropicvertex`. It is priced on its own, so it needs its own rates on the **Model Pricing** tab. On that tab its name starts with the provider, for example `anthropicvertex/claude-sonnet-4-6`.
### **Set model rates on the Model Pricing tab**
All rates on this tab are in US dollars per million tokens. Each model has four rates, one for each kind of token count described in [How cost is worked out](#how-cost-is-worked-out).
#### **Models used but not priced**
The box at the top lists every model Kaiya used in the last 30 days that has no rates yet, with a count of how many there are. Add their rates so their usage is costed in full. The box is the quickest way to find gaps in your cost report.
Select a model's chip to add it to the pricing table. The model is added and saved straight away, with all four rates **Not set** and a **Custom** badge, and it leaves the list. Enter its rates next.
{% hint style="info" %}
A chip uses the exact name the cost report expects, including the provider for a model used through another provider. Adding a model from its chip avoids a typing mistake that would leave the usage unpriced.
{% endhint %}
#### **Add a model by name**
You need the Super User role to add a model.
1. Go to **Settings → Intelligence → Kaiya & AI → Kaiya Cost** and open the **Model Pricing** tab.
2. Type the model's name under **Add a model**, exactly as it appears in **Cost by model**.
3. Select **Add pricing**. The model is added to the pricing table.
4. Enter the four rates in **Input $/1M**, **Cached input $/1M**, **Cache write $/1M** and **Output $/1M**.
5. Add any other names for the same model under **Synonyms**, if you need them. See [Other names for a model](#other-names-for-a-model).
6. Save the row.
A row that has no unsaved changes shows **Saved**.
#### **Change or delete a model's rates**
Super Users can edit any rate in place, then save the row. Select **Delete** on a row to remove the model from the pricing table.
#### **Other names for a model**
One model can be known by more than one name, such as a dated snapshot of the model. Add these as synonyms of the model, and they share its rates. Each name can belong to one model only. Tellius doesn't accept a name that another model already uses.
### **Cap spending on the Limits tab**
The **Limits** tab sets monthly spending caps in US dollars. Each cap is checked against cost for the current calendar month to date, so caps start again on the first day of each month. An empty cap, shown as **No cap**, means no limit.
* **Instance limit** caps the total spend of everyone in the workspace. When the workspace's month-to-date cost reaches it, Kaiya stops taking new requests from anyone until the cap is raised or a new month starts.
* **Default per-user limit** is the cap for every user who doesn't have a cap of their own. Use it to give everyone the same monthly allowance in one place.
* **Per-user limits** lists every user with their month-to-date spend under **Used (MTD)**. Search for a user by name or email. Enter an amount under **Monthly cap ($)** to give that user their own cap in place of the default, and save the row. Clear it to put the user back on the default. **Usage** and **Status** show where each user stands against their cap.
{% hint style="warning" %}
A cap of $0 blocks the user straight away. To remove a user's limit, clear the cap rather than entering 0.
{% endhint %}
Both kinds of cap apply together. A user can't send new Kaiya requests once either their own cap or the instance cap is reached.
#### **What users see as they near a cap**
* **At 80% of a cap,** an amber banner warns the user that they are close to the limit.
* **At 100% of a cap,** a red banner appears. The user can't send messages to Kaiya or export until the cap is raised or a new month starts.
Kaiya Cost doesn't send email alerts. There is no custom billing period: caps always count the current calendar month to date.
### **Best practices**
* **Check for unpriced models after a model change.** When you add a model connection, open **Model Pricing** and look under models used but not priced. Add their rates so their usage shows a cost.
* **Add models from their chips.** The chip carries the exact name the report expects, including any provider.
* **Use synonyms for model snapshots.** When a new name is the same model as one already priced, add it as a synonym rather than a new model. The rates then live in one place.
* **Set a default per-user limit before individual caps.** Give most users the same allowance, then set your own caps only for the people who need more or less.
### **Related pages**
* [LLM Configuration](https://help.tellius.com/settings/kaiya-and-ai/llm-configuration)
* [LLM Assignments](https://help.tellius.com/settings/kaiya-and-ai/llm-assignments)
* [Setting up LLM for Kaiya](https://help.tellius.com/settings/application-settings/setting-up-llm-for-kaiya)
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://help.tellius.com/tellius-6.4/settings/kaiya-and-ai/kaiya-cost.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
