OpenAI prompt-cache retention

OpenAI's platform changelog lists a default 24-hour prompt-cache retention behavior for organizations that are not using zero data retention. Why it matters: Agent cost and latency assumptions increasingly depend on cache behavior, hosted tools, and data-retention settings rather than only per-token model pricing.

Try this: Before scaling a repeat agent task, check whether prompts are cacheable, what retention policy applies, and whether the workflow requires a stricter data setting.