OpenAI Adds a 12-Hour Prompt Caching Option to the Responses API, With Cache Writes Priced at 2.5 Times the Input Rate

Overview

An October 20 entry in the OpenAI API changelog says: “You can now choose 12-hour prompt caching in the API.” The entry is tagged to the v1/responses endpoint. According to the same entry, the option “keeps shared instructions, tools, and context cached longer to benefit from lower cached-input pricing, even when requests are hours apart.”

What We Know

• Price. The changelog says “12-hour cache writes cost 2.5x the input token price.” OpenAI’s prompt caching guide lists a 30-minute lifetime, the default, at 1.25 times the uncached input-token rate and a 12-hour lifetime at 2.5 times that rate. The guide adds: “These are total write rates, not surcharges on top of ordinary input.”
• Reads. The guide says cache reads cost 0.1 times the ordinary input rate on most GPT-5.6-and-later models and 0.05 times on GPT-6.1 Sol, regardless of the selected lifetime.
• How to choose it. For models that support it, the guide says the lifetime is set with the prompt_cache_options.ttl request parameter, with values "30m" (default) or "12h" in the Responses API. A prompt_cache_breakpoint.ttl setting overrides the request-level value for an individual breakpoint, so one request can mix lifetimes.
• Billing rules. According to the guide, upgrading valid 30-minute coverage to 12 hours “charges a new 12-hour write for those tokens, without credit for the earlier write.” Any portion already covered by a valid 12-hour entry remains a cache read.
• Usage reporting. The guide says input_tokens_details.cache_write_tokens_details breaks writes into ttl_30m_tokens and ttl_12h_tokens, and that the two counts do not overlap.
• Worked example. The guide prices a request at $1 per million input tokens and $0.10 per million cached-input tokens, with 1,000 ordinary tokens, 4,000 cached reads, 3,000 30-minute write tokens and 4,000 12-hour write tokens. It gives the input cost as $0.01515.
• Guidance. The guide suggests considering "12h" for stable prefixes reused across longer gaps and says: “Compare the higher 12-hour write rate with the cost of rewriting the prefix after 30-minute coverage expires.”

What We Don’t Know

• The guide says OpenAI “may retain entries longer than the requested minimum” and that "12h" “does not set a deletion deadline or change the maximum data-retention policy.” The 12-hour figure is therefore a minimum lifetime, not a guaranteed expiry.
• The guide also says the minimum lifetime governs eligibility for cache-read pricing and “does not guarantee that every request reuses stored KV tensors or has lower latency.”
• Both sources are OpenAI’s own documentation. No independent measurement of savings from the 12-hour option was reviewed for this article, and neither page gives a per-million-token price for the 12-hour write, which is stated only as a multiple of each model’s input price.