Service update · OpenAI
OpenAI API adds a 12-hour prompt caching option
- Review updated
OpenAI announced 12-hour prompt caching for its API on October 20, 2026, at 17:56 UTC (01:56 Beijing time on October 21). Developers can reuse shared instructions, tools and context at the lower cached-input rate even when requests are hours apart. They can retain the 30-minute default or combine both durations in one request, using 12 hours for shared instructions and 30 minutes for newer conversation content. Longer caching carries an extra cost; OpenAI says repeated reuse throughout the day can more than offset it. The announcement does not specify pricing.