OpenAI’s 28-Day Product Sprint Day 16: Codex and Work Caches Permanently Retained, APIs Extended to 12 Hours

Insight Beating AI Flash News: OpenAI’s Product and Platform Lead Thibault Sottiaux announced Day 16 of the company’s 28-day product sprint, with a key focus on resolving context caching issues for long-running Agents. For logged-in Codex and ChatGPT Work users, OpenAI has announced permanent retention of prompt caches. This ensures cache hits even when resuming work after hours or reopening old sessions, eliminating the loss of cache reuse benefits due to long pauses. The API has also been updated: OpenAI extended the optional cache duration for the Responses API from the default 30 minutes to 12 hours. Developers can cache system instructions, tool definitions, and common context long-term, so Agents can reuse prior computation results even when resuming tasks hours later. The two durations can be mixed: fixed instructions are retained for 12 hours, while new conversation content is cached for only 30 minutes. However, extending API caching incurs additional costs. The initial write price for 12-hour caching is 2.5 times that of standard input, double the cost of the default 30-minute cache. For subsequent cache hits, input prices for most new models are reduced by 90%, and by 95% for GPT-6.1 Sol. Thus, long caching is more suitable for Agents that repeatedly call the same context, rather than short, one-off tasks.