Skip to main content
Track your Braintrust usage and costs on the Billing page. Usage analytics and Inference activity are available on all plans. Pro plans and Starter plans with on-demand usage enabled also show invoice projections and a Detailed usage breakdown link to the Orb portal. For Enterprise invoice details, contact Braintrust.

​
View usage

  1. Go to Settings > Billing.
  2. If available, review On-demand usage for your projected invoice, Logs and Scores breakdown, and billing cycle reset date.
  3. Under Usage analytics, select the current month or a custom date range (up to 100 days). Use the scope dropdown to view Entire organization (default) or specific projects. To compare every project, click Select all in the dropdown, or click Deselect all to return to Entire organization. Project comparisons support up to 200 projects, so if you can access more than 200, these actions are hidden and you select a subset instead.
    • Entire organization: toggle between Daily and Cumulative views of logs (GB), scores, and built-in inference tokens.
    • By project: daily logs, scores, and built-in inference tokens broken down by project. Project data is available from September 1, 2026, and selecting a start date before then prompts you to choose a later date.
    In either scope, the built-in inference chart covers built-in model and Topics tokens. Use its dropdown to break usage down By model, By token kind (input, cached input, cached write, and output), or both. On every plan except Starter without on-demand usage, use the mode dropdown to switch all charts between Usage and Cost. The dropdown shows the current mode. On Enterprise plans, cost mode:
    • Shows list-rate estimates for inference and Topics instead of your contract price, labeled Estimated cost in chart subtitles.
    • Hides the logs and scores charts, because your contract sets their pricing.
    To view a chart at a larger size, hover over it and click . In either scope, click CSV to download the data for the selected date range.
  4. Review Inference activity to see how your Braintrust Gateway token usage is distributed over the last 7 days. Each daily line chart shows each item’s share of that day’s tokens, and the legend shows its share of the whole period. Hover over the icon by a chart title for definitions of its “Other” and “Unknown” items. This section appears only if you have Read permission in the All projects column.
    • Activity by feature: share by Braintrust feature (Online scoring, Topics, Playground, Experiments, and Loop). Other feature includes requests from sources Braintrust doesn’t recognize, and Unknown feature includes requests with no feature label, such as your own SDK and API calls through Gateway. Use the All projects dropdown to show the feature mix of a single project.
    • Activity by project: share for your top five projects. Other projects includes the remaining projects, deleted projects, and projects you can’t access. Unknown project includes requests that weren’t associated with a project.
  5. If available, click Detailed usage breakdown under On-demand usage to open the Orb portal, where you can:
    • View detailed breakdowns of trace spans, processed data, scores, Topics tokens, and retention storage.
    • Download usage reports for internal tracking or accounting.
    • Monitor costs and billing cycles.
    • Compare usage across multiple billing periods.

​
Understanding usage metrics

Your Braintrust usage is measured across four dimensions. Each contributes to your monthly costs if your usage grows beyond your plan’s included limits. See Plans and limits for included limits and on-demand usage rates by plan, and Model credits for the monthly model credit and overage rates.

​
Processed data

Processed data refers to the total bytes of data ingested by Braintrust when you create logs or experiments. This includes:
  • Inputs and outputs
  • Prompts and completions
  • Metadata and tags
  • Traces and spans
  • Datasets
  • Attachments (images, audio, files)
  • Any other related information
How it’s measured: Braintrust measures the size of each log, experiment, and dataset record when you send it and adds it to your total for that calendar month, which is invoiced at the end of the month. On Enterprise plans, monthly usage draws down from your committed volume.
Processed data is measured at ingestion, based on what you send, not on what’s currently stored. Deleting data or shortening your retention window doesn’t reduce your processed data usage for the month. Retention storage, a separate charge on Pro and Enterprise plans, is measured the same way, so deletion and shorter windows don’t reduce it either.

​
Scores

Scores are used to measure the results of offline or online evaluations run in Braintrust. Each time you record a score, it counts toward your monthly usage. Examples of scored operations:
  • Offline evaluation scores from experiments
  • Online evaluation scores from production logs
  • Human review scores
  • Custom scorer outputs
How it’s counted: Each score recorded (whether automatic or manual) increments your monthly score count by one.

​
Topics tokens

Topics is metered in input and output tokens consumed by the daily Topics pipeline (facet summarization, embedding, and cluster naming). Each plan includes a monthly credit, shared with the built-in models, that covers a portion of token usage at no extra cost. Once the credit is used up, uniform overage rates apply across all plans. How it’s measured: Input and output tokens are summed daily across all Topics pipeline calls and counted toward your monthly credit. The credit does not roll over to the next month.
On the Starter plan, if you exhaust your monthly credit and haven’t enabled on-demand usage, Topics is paused for the rest of the billing cycle. Add a card to continue running Topics at the overage rates, or wait for the next monthly cycle.
Built-in model usage also draws from your model credits and appears under Native inference in your usage breakdown. Topics and built-in model tokens are charted together in the built-in inference chart under Usage analytics.

​
Retention storage

On the Pro plan, retaining data beyond 30-day retention incurs a retention storage charge. Like processed data, retention storage is measured at ingestion: it bills the data you ingest into each month you keep beyond the base window, not what’s currently stored. This covers logs and playground data, including each row’s activity history. Experiments aren’t subject to your retention window and have up to 365-day retention on all plans, but the data you ingest into them still counts toward retention storage. On Enterprise plans, retention beyond your base window is arranged with your account team. To keep data longer, extend your retention window. Because retention storage is measured at ingestion, deleting data, setting a retention automation, or shortening your window doesn’t reduce it. The only way to lower retention storage is to ingest less. How it’s measured: Braintrust measures the size of your retained data daily and bills the highest daily amount during the billing month. See Plans and limits and the Pricing page for rates.

​
Set up spend alerts

Spend alerts are available for plans with on-demand usage (Starter and Pro).
Receive a notification when your total invoice amount reaches a threshold in the current billing cycle. You can set up to three spending thresholds and choose to notify via Billing email, a Slack channel, or both.
  1. Go to Settings > Billing.
  2. Click Spend alerts.
  3. Set your spending thresholds.
  4. Under Notify via, select Billing email, Slack channel, or both.
  5. Click Save.
Spend alerts are informational-only and do not interrupt or limit your usage. You can update or disable alerts at any time. Spend alerts are distinct from Topics credit notifications. Braintrust emails you automatically when you’ve consumed 60% and 100% of your monthly Topics credit. These notifications are sent automatically and aren’t configurable, whereas spend alerts track your total invoice amount against thresholds you set.

​
Next steps