50 prompts in your current set.
More repetitions = more reliable results, but higher cost.
Estimate uses public per-token pricing. Final cost is recorded after the run.