The update aims to reduce cost spikes for long running and ambient agents that previously suffered from frequent cache misses. By extending the duration of stored context, OpenAI minimizes the need for the model to reprocess large amounts of history during repeated interactions.
Wednesday, October 21, 2026
Story
OpenAI Launches Permanent Prompt Cache for Codex and Work
AI8 hours ago
Developers using the OpenAI API can now select a 12 hour prompt caching window to maintain context for persistent agents such as dots throughout the day. This allows for the reuse of shared instructions and tools at a lower cached input rate even when requests are hours apart. For users logged into Codex and Work, the company has implemented a permanent cache that eliminates penalties for resuming old conversations.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.