Usage and spend
See how much your organization used over time: model requests, tokens, spend and compute. Open Usage from the left rail.
The period
The dates at the top set the period every chart covers: one month ending now, unless you pick other dates. While it ends now, new usage appears each time you open the page or select Refresh; Last month brings back the default. Choose whether charts count by Day, Hour or Minute.
What the page shows
| Section | Charts |
|---|---|
| Overview | Model requests, and failed attempts. |
| Tokens | Input tokens, output tokens, and requests per model. |
| Cost | Spend at your provider's prices (estimated), and what your plan has used this billing period. |
| Compute | How long runs took, and the CPU and memory the steps' code used while it ran. |
Each chart
The total for the period is shown above each chart. Its ⚙ menu splits it by model, provider, project, workflow or status; draws it as bars, lines or areas; makes it cumulative or stacked; and, where it applies, changes what it measures, for example the error rate instead of the error count. Download saves the chart's numbers as a CSV.
Where the cost comes from
With your own key, you pay the provider for tokens; the spend shown is estimated from each model's token prices. Each run also shows its own tokens and cost live in the DAG tab, and each step in its console's Cost tab. Patterns - where spend goes, which steps fail, when usage peaks - are on Insights.