What will your AI feature cost?
Estimate a real workload, compare published model rates, and keep a useful shortlist.
Start with traffic, not a monthly guess
The Cost planner takes input tokens, output tokens, requests per day, and planning days. These are your assumptions, not recorded account usage. Try a quiet day and a busy day to see how traffic changes the estimate.
Compare the same workload
Select a model from the live catalog. The planner uses its separate input and output rates and shows other models with the lowest published token costs for the same inputs. Price does not measure quality, reliability, or capacity. Test the models on your own tasks before choosing.
Save candidates while you decide
Open a model in the catalog and choose Save model. The Saved models filter keeps those candidates together. Your shortlist stays in this browser and does not sync to your account or other devices. Copy the estimate when you want to discuss a budget.
Reconcile estimates with actual usage
The planner estimates gross text-token charges. Allowances and credits can reduce what you pay; reasoning, images, taxes, and provider accounting can change the final amount. Review Usage and Request logs after a small real test, then update your assumptions.