OpenAI API Launches 12-Hour Prompt Caching; Cache for Codex and Work Logged-In Users Retained Permanently

October 21 — According to OpenAI Developers, the OpenAI API now supports 12-hour prompt caching. Developers can cache shared instructions, tools, and context for longer, and even if requests are hours apart, they can still be billed at the lower cached input price. This feature is mainly aimed at agents that run continuously and reuse context all day. Tibo (@thsottiaux) later added that the cache for logged-in users of Codex and Work will be retained permanently. When users return to an old conversation, or continue it hours later, they will hit the cache without incurring additional costs.