There is more than one cost component
The Agents API overview points to the selected model’s API rates, standard OpenAI tool rates, and standard hosted container rates. Your own infrastructure and third-party tools can add separate costs. Read current official pricing for the exact model and tool configuration.
Estimate per successful task
Measure input and output tokens across turns, tool calls, and sandbox usage. Divide observed spend by successfully completed tasks. Retries and subagents can change this figure substantially even when the end-user prompt looks identical.
Reduce unnecessary work
Use a suitable model, bounded tasks, narrow tool outputs, and only the environment capabilities required. Compare cache behavior and repeated context overhead using real usage records. Do not assume adding concurrency lowers total cost.
Budget with explicit assumptions
The calculator on this site is a token-only estimate with rates you supply. It excludes tools, containers, third-party services, taxes, and application hosting. It is an arithmetic planning aid, not a billing quote.
Estimate token cost
Enter rates from the official pricing page. Excludes tools, containers, hosting, and taxes.
Read the official reference
Check the source for current API fields, account requirements, and service limits.
OpenAI: overview Current model and tool pricing