Understand usage and estimated cost
Read reported token usage beside the task, its retries and its review outcome. Check the price basis before treating a usage figure as cost.
Estimates depend on reported usage, model identity and available historical prices. Unknown prices remain unknown. Compare total effort for an accepted result, including human review.
Count the whole task
Include retries, failed attempts and follow-up runs. Record human review and verification effort separately from token usage so an inexpensive run does not hide an expensive task.
Read usage with its outcome
Inspect the task result, then its runs and reported usage. Check why rework was needed: missing context, unclear scope or insufficient evidence. Compare models on similar tasks and acceptance criteria.
Historical seven-day usage example from July 2026, grouped by project and operating mode. These are captured values, not the current workspace total.
Check the price basis
The Studio pricing adapter uses Token Economy to look up dated prices for the model and run timestamp. An estimate is separate from a provider invoice or subscription charge. Missing prices must remain visible when reading totals.
Compare accepted outcomes
Use the same expected behavior and checks when comparing workflows. Include failures, unresolved findings and review time. Repeat the comparison before generalizing from one task.