OpenAI's GPT‑6 prompt caching lifts hit rates by default

GPT‑6 is designed to support persistent agents that work for hours on complex tasks, from refactoring codebases to producing well-researched documents. Those agents make a series of API requests that build on each other, often carrying forward the same instructions, tool definitions, and context. OpenAI caches that shared context to reuse computation across requests, reducing response times and giving developers discounts on cached input tokens.

The new prompt caching system builds on that engine by delivering higher hit rates by default and adding tools to help developers see and fix misses. A dashboard shows how much of an application's input is served from cache, and a diagnostics tool compares a request with a recent response to identify what changed. Developers can also choose explicit cache breakpoints, adjust reasoning effort without breaking cache, keep tool definitions stable, and prewarm the cache.

The value of these controls hinges on how well developers manage stability in their prompts and tools. Teams that keep their setups consistent may capture more of the default performance gains, while those with frequent changes may need the new diagnostics to understand where reuse is lost.