> ## Documentation Index
> Fetch the complete documentation index at: https://docs.terma.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> Compare the models your agents use by calls, tokens, spend, sessions, and people, put up to four side by side, and see each model's list prices.

The **Models** pages show which models your agents use, how much, and at what price. **Models → Usage** lists them; **Models → Compare** puts them side by side.

## Usage

Search for a model, or sort by **Highest spend** (or **Most tokens**), **Most calls**, **Most sessions**, **Most people**, or **Model name**.

| Column | Description |
| - | - |
| **Model** | The model, linking to its page |
| **Calls** | Model calls in the period |
| **Input tokens**, **Output tokens** | Tokens sent and received |
| **Spend** | Spend in the current usage view (or input plus output tokens in the Tokens view) |
| **Sessions** | Sessions that used the model |
| **People** | People who used it; select the count to see who |
| **Tool calls** | Tool calls made in its turns |
| **Spend share** / **Token share** | Its share of the project's total |

Tick the **Compare** checkbox on up to four models to compare them.

## Compare

Compare puts up to four models side by side. Pick them in the bar at the top of the page, or from the Usage list.

Daily charts show each model's spend, tokens, model calls, people, share served from cache, time to first token, and failed calls. Below them, a table compares:

* **What it burnt** — tokens and model calls
* **The work in its turns** — sessions, tool calls, compactions, and subagents
* **Per unit** — cost per million tokens, share served from cache, output per call, time to first token, and failed calls
* **List price, per 1M tokens** — input, output, and cache prices, with a pop-up that ranks providers
* **Where the tokens go** — the mix of input, output, cache read, and cache write
* **Who uses it** — the people who called it

Rows mark which model is **Better** or **Cheapest**, and show ratios such as **3×** or **about the same**.

## A model's page

Select a model to open its page:

* **Called through** — the sources that used it
* **Spend**, **Tokens**, **Sessions**, **Active users**, and **Model calls**, with spend over time
* **Sources** and **Providers** — select one to narrow the page to it — plus principals and sessions
* **Price history** — input, output, cache read, and cache write prices over time
* **Current list prices** — per provider, per million tokens: input, output, cache read, cache write, and combined, with when each was observed

## From the CLI

```bash theme={null}
terma usage --group-by model --since 30d
terma usage --group-by provider --since 7d
terma session list --model claude-opus-4-1 --sort cost
```