OpenAI adds optional 12-hour prompt caching to its API
What happened
On October 21, OpenAI's developer account said the API now supports optional 12-hour prompt caching. Long-running agents can reuse earlier instructions, tools and context even when requests are hours apart, billed at the lower cached-input rate. Tibo later added that logged-in users of Codex and Work keep their cache permanently: returning to an old conversation or resuming a session hours later is guaranteed a cache hit, with no extra penalty.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- AI HOT · ProductsOpenAI API 支持 12 小时提示词缓存，Codex 与 Work 登录用户缓存永久保留
OpenAI 开发者账号宣布，API 现已支持可选的 12 小时提示词缓存。持续运行的智能体可在请求间隔数小时后，以更低的缓存输入费率复用指令、工具和上下文。Tibo 补充称，Codex 与 Work 中已登录用户的缓存会永久保留，返回旧对话或数小时后恢复会话保证命中缓存，且无额外惩罚。
Heat over time
Not enough continuous observations to draw a trend yet.