Event ·Official first-hand

GPT-6 改进提示词缓存：更高命中率与显式断点控制

OpenAI 为 GPT-6 引入了更完善的 prompt caching 机制，官方称缓存命中率较此前更高。新增 diagnostics 工具可查看缓存命中情况，方便排查延迟问题。开发者还能使用 explicit breakpoints 手动标记缓存边界，并调整缓存控制选项。这些改动同时降低了推理延迟和 token 费用，对长上下文或多轮对话应用尤其有用。

Current read

OpenAI 给 GPT-6 的提示词缓存加了诊断工具和手动断点，写长上下文应用的人能直接省 token 费用，还能自己控制缓存边界。

11 sources37° AI heatLast updated 2026/9/22 21:00:00

Evidence

11 sources

Primary sourceOpenAI Blog

2026/9/22 21:00:00View source

Related source 1@OpenAIDevs

2026/9/22 21:13:01View source

Related source 2lmarena.ai

2026/9/22 19:29:01View source

Related source 3Lovable

2026/9/22 18:08:24View source

Related source 4OpenRouter

2026/9/22 18:23:35View source

Related source 5Sam Altman

2026/9/22 18:26:54View source

Related source 6GitHub

2026/9/22 18:41:29View source

Related source 7Simon Willison’s Weblog

2026/9/22 18:48:37View source

Related source 8AI SDK

2026/9/22 19:18:38View source

Related source 9Decoder

2026/9/22 20:06:51View source

Related source 10Blognone

2026/9/23 01:22:58View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story