Skip to main content
For the complete documentation index for agents and LLMs, see llms.txt.

Understand LLM CostsBeta

Beta
This is coming — this page describes a feature that's still being rolled out and may not be available to you yet.

To track how much your organization spends on LLM calls and find out which workspaces, pipelines, or models drive that spend, use the Cost Management dashboard. It estimates your LLM spend from traced token usage and per-model prices. The Usage dashboard is separate: it shows credits and document storage units.

Permissions

You must be an Organization Admin, or have the costs read permission, to view the Cost Management dashboard.

Open the Dashboard​

Go to Operate>Cost Management in the sidebar.

Cost Summary Tiles​

To check your overall spend and whether it's on track for the month, start with the five tiles at the top of the dashboard. They summarize your organization's LLM spend for the selected date range. The first four also show the percentage change compared to the previous period:

  • Total cost: The total estimated cost, in dollars, of all LLM calls.
  • Total requests: The number of LLM calls made.
  • Total tokens: The number of tokens consumed across all LLM calls.
  • Avg. cost / request: The average cost per LLM call.
  • Projected this month: Your estimated total spend for the current calendar month: month-to-date spend plus the trailing seven-day daily average for each remaining day.

Cost Over Time​

The Cost over time chart plots your daily estimated spend across the selected date range so you can spot spikes or trends.

Cost Breakdown Charts and Tables​

To find out which providers, models, workspaces, pipelines, or deployments drive your spend, use the breakdown charts and tables below the summary tiles:

  • Cost by provider: A bar chart showing your top LLM providers by cost.
  • Cost by model: A bar chart showing your top models by cost.
  • Cost by workspace: A paginated table showing cost, requests, and tokens per workspace.
  • Cost by pipeline: A paginated table showing cost, requests, and tokens per pipeline.
  • Cost by deployment: A paginated table showing cost, requests, and tokens per deployment.

Click a column header to sort a table by that column. Click a bar in the provider or model chart, or a row in a breakdown table, to filter the whole dashboard down to that item.

Cost Flow​

To see which workspaces and deployments account for the spend on each model, use the Cost flow chart. It shows how your spend moves from your organization, through workspaces and deployments, to individual models. Hover over a connection to see the estimated cost it represents.

Filters​

To focus on one part of your spend, use the Filters panel. You can filter by workspace, pipeline, provider, model, deployment, and API key. Active filters appear as chips at the top of the dashboard. Click Clear filters to remove all active filters at once.

When you click a bar, row, or node, the dashboard heading changes to Viewing [dimension]: [name]. Click Back to overview to return to the organization-wide view.

Incomplete Cost Data​

If some of your LLM usage can't be fully reflected in the totals yet, a warning banner appears above the summary tiles. This can happen when:

  • Some calls use a model that doesn't have a price configured yet (unpriced).
  • Some calls are still being priced and update automatically (pending).

To find out how much usage the totals don't reflect yet, click Show details on the warning. It shows how many calls and tokens are affected, and which models are missing pricing.