Skip to content

What does Fab's context cost include?

Context cost estimates the value of the context tokens delivered by Cortex Flashes in the selected period. Fab applies its stated standard input-token rate. The figure covers retrieved knowledge; it does not measure your total model spend or your subscription bill.

Open the dashboard options menu and select About this data to see the model, rate and dated rate card used for the estimate. Use the displayed rate when interpreting a report, because a price assumption is part of that report's basis.

Context tokens / shipped Issue divides the period's Flash context tokens by the number of Issues shipped in the period. Fab counts an Issue as shipped when it is marked done. Context cost / shipped Issue divides the corresponding estimate by the same count. These are period averages. They do not show how much each individual Issue consumed. Context spent on canceled or unfinished work stays in the totals, while canceled Issues are excluded from the count.

When a Module is selected, context totals count Flashes that targeted that Module. A Module included as a dependency is different from a direct Flash request. One explicit request counts once even when it includes several dependencies.

If no Issues were shipped, the per-Issue figures are unavailable. Fab can still show context tokens and context cost for that period.

Read the cost estimate alongside throughput, the lifecycle no-reopen proxy, recorded candidate acceptance and coverage. More retrieved context does not by itself show that an agent used it correctly or that the work improved.