Token Gremlin(@TokenGremlin):Day 16: OpenAI just extended prompt caching in the API from the usual 30 minutes to an optional 12 hours. This means developers can reuse instructions, tools, and other context that a model has already processed, even when requests are hours apart, without having to process everything from scratch again. For Codex and Work, Tibo says the benefit goes even further. You can return to an old conversation after several hours and still get a cache hit. The 12-hour option costs more upfront, but repeated use of the same context can make it cheaper overall.

2026.10.21 01:22

Day 16: OpenAI just extended prompt caching in the API from the usual 30 minutes to an optional 12 hours. This means developers can reuse instructions, tools, and other context that a model has already processed, even when requests are hours apart, without having to process everything from scratch again. For Codex and Work, Tibo says the benefit goes even further. You can return to an old conversation after several hours and still get a cache hit. The 12-hour option costs more upfront, but repeated use of the same context can make it cheaper overall.

Tibo@thsottiaux

2026.10.21 01:10

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.