OpenAI Extends API Context Caching Duration for Autonomous Agents

OpenAI announced a new feature for its developer platform that allows software creators to store system instructions and background memory for up to twelve hours. Previously, this stored memory, known as prompt caching, expired after thirty minutes by default.

For everyday users, this technical update matters because it lowers the cost of running digital assistants that work continuously in the background. Software developers can now keep complex tools and guidelines active all day without paying full price to resend that information whenever an agent acts hours later.

According to posts from OpenAI's official developer channel on X, creators can also mix retention times within a single request. They can assign the twelve-hour window to permanent guidelines while letting recent conversation history refresh on the standard thirty-minute timeline.

While this makes running persistent AI assistants far more affordable, it raises questions about how competing model providers will adapt. It remains unclear whether offering longer memory storage will become the standard pricing model across the industry as autonomous software tools become more common.