OpenAI has outlined improvements to prompt caching in GPT-6, describing a set of changes intended to make repeated prompts cheaper and faster to process.
According to the company, the update delivers higher cache hit rates, meaning more repeated prompt content can be served from cache rather than reprocessed. OpenAI also points to new diagnostics that give developers more visibility into how caching behaves.
The release adds explicit breakpoints and additional controls, which OpenAI says are designed to reduce latency and costs. The company frames these as practical levers for teams building on GPT-6.
OpenAI did not disclose specific benchmark figures, pricing, or availability dates in the announcement.