OpenAI – 12 hour cache (API) available and "cache forever" for work/codex users

OpenAI now offers a 12-hour prompt caching option in its API, announced by Tibo (OpenAI staff) on X as "Day 16" of a series, and keeps cache indefinitely for logged-in users across Codex and Work with no penalty for resuming an old conversation after a few hours. OpenAI said developers can keep shared instructions, tools and context cached longer and pay the lower cached-input rate even when requests are hours apart, a setup it said suits persistent agents that reuse context throughout the day.

Tibo on X: "Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache." / X Tibo on X: "Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache." Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache. You can now choose 12-hour prompt caching in the API. Persistent agents, like dots, reuse context throughout the day. You can keep shared instructions, tools, and context cached longer and pay the lower cached-input rate, even when requests are hours apart. Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache. You can now choose 12-hour prompt caching in the API. Persistent agents, like dots, reuse context throughout the day. You can keep shared instructions, tools, and context cached longer and pay the lower cached-input rate, even when requests are hours apart.