Permanently retain cache! OpenAI's major update, longer than DeepSeek?

What happened: OpenAI's API will offer a 12 hour cache retention option, and Codex and Work logged in users will get permanent cache retention, per ChatGPT and Codex lead Tibo. Key numbers: 12 hour cache option for GPT-5.6 and above models; default is 30 minutes. Why it matters: OpenAI's docs state the 12 hour setting is a minimum lifetime for cache read billing eligibility and does not guarantee reuse or lower latency. The update follows DeepSeek open sourcing a reasoning framework with cache retention designed for up to 3 days.

OpenAI's API will offer a 12 hour cache retention option, and Codex and Work logged in users will get permanent cache retention, per ChatGPT and Codex lead Tibo.

OpenAI's docs state the 12 hour setting is a minimum lifetime for cache read billing eligibility and does not guarantee reuse or lower latency. The update follows DeepSeek open sourcing a reasoning framework with cache retention designed for up to 3 days.