# Usage

> Compare what you spent by model or key, and export the current view.

Open [Usage](https://deference.si/usage) to see what your requests cost over time. To inspect a single request, use [Activity](https://deference.si/docs/dashboard/activity).

## Choose what to compare

1. Choose 24 h, 7 d, 30 d, 90 d or Month to date. The page opens on **30 d**, or on **7 d** when every request in the last 30 days falls in the last 7.
2. Select **Spend**, **Requests**, **Tokens** or **Avg latency** above the chart. Latency appears when the range has timing data.
3. Use **Group by** to compare by **Model**, **Key** or **Client**.

Spend is credit charged, tokens include input and output, and average latency measures time to the end of the response. Each total also shows its change from the previous period.

The chart groups requests by UTC day, or by hour for the 24 h view. It shows your top three series and combines the rest into Other. For **Spend**, **Requests** and **Tokens**, hover a bar to read its values and select a legend entry to hide a series. **Avg latency** draws a line.

## Find what used your credit

The table below the chart follows the **Group by** choice. Each row shows requests, input and output tokens, spend and its share of spend. Select a row to filter the totals, the chart and the table to that model, key or client. When there are more than ten rows, **Show all \{count}** lists the rest and **Show top 10** returns to the first ten.

The range, grouping, metric and filters stay in the URL. Bookmark it to return to the same view. Another account opening the URL sees its own usage.

## Export the figures

Select **Export CSV** to download one row per period and group for the chosen range, grouping and filters.

## When the view is empty

Usage shows **No usage yet** when there are no requests in the last 90 days. If you have requests outside the selected range, choose **Show last 90 days** or change the range.

Use D, W, M, Q and T to switch to 24 h, 7 d, 30 d, 90 d and Month to date.