OpenAI has added a 12-hour prompt caching option to its API, letting developers keep shared instructions, tools and context cached long enough that requests made hours apart can still be billed at the lower cached-input rate. The feature, listed in OpenAI's API changelog on October 20, applies to the Responses API.
The trade-off is a higher write price. Under OpenAI's prompt caching guide, cache writes at the 12-hour setting cost 2.5 times the uncached input rate, compared with 1.25 times for the default 30-minute setting. Cache reads cost 0.1 times the ordinary input rate, or 0.05 times on GPT-6.1 Sol. The guide specifies that cache-write pricing is not an additive fee.
Developers pick the lifetime with the prompt_cache_options.ttl parameter, which accepts "30m" or "12h". A separate prompt_cache_breakpoint.ttl setting can override the lifetime for an individual explicit breakpoint.
Key facts
- 01OpenAI added a 12-hour prompt caching option to the Responses API on October 20, 2026.
- 0212-hour cache writes cost 2.5x the uncached input rate, versus 1.25x for the default 30-minute cache.
- 03Cache reads bill at 0.1x the ordinary input rate, or 0.05x on GPT-6.1 Sol.
- 04The option is set with prompt_cache_options.ttl and is limited to GPT-5.6 and later models.
- 05The minimum cacheable prompt for GPT-5.6 and later is 1,024 tokens.
Both values set a minimum lifetime rather than a deletion deadline. OpenAI says it may keep cache entries longer than the minimum.
The 12-hour option is limited to GPT-5.6 and later models in the Responses API. The minimum cacheable prompt for those models is 1,024 tokens. OpenAI's guide recommends the longer setting for stable instructions or reference material reused across longer gaps.
Older models use a different control, prompt_cache_retention, with in_memory and 24h values. In-memory retention keeps entries roughly five to 10 minutes after last use, up to one hour, while the 24h option can hold them for up to a day on models including GPT-5.5, GPT-5.4 and GPT-4.1.
The 12-hour cache extends a run of caching tools OpenAI has shipped this year. A Prompt Caching dashboard tracking hit rate and reads per write arrived on the API platform on August 20, and Prompt Cache Diagnostics, which compares cache reuse against a previous response to explain misses, became generally available in the Responses API on September 8.
Earlier, on May 29, OpenAI changed the default prompt_cache_retention value to 24h, instead of in_memory, for organizations without zero data retention enabled.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
