OpenAI adds optional 12-hour prompt caching to the Responses API
In short: OpenAI's Responses API now supports a 12-hour prompt cache option, keeping shared instructions, tools and context cached for longer so repeated requests hours apart can still benefit from cheaper cached-input pricing. Writing to this longer cache costs 2.5x the standard input token price. This is an optional setting alongside the existing shorter default cache.
This summary was generated automatically by AI from OpenAI API Changelog's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1New optional 12-hour prompt cache for v1/responses
- 2Cached context (instructions, tools) stays available even when requests are hours apart
- 312-hour cache writes cost 2.5x the standard input token price
| Parameter | Before | Now |
|---|---|---|
| Prompt cache duration | Not specified | 12 hours (optional) |
| 12-hour cache write price | Not specified | 2.5x input token price |
Why it matters
Applications that send large, repeated system prompts or tool schemas but don't call the API constantly can now avoid re-paying full input price on every request, as long as the time gap stays under 12 hours.
What it means for AI agents and contact centers
For voice agent and contact-center integrations that reuse long system prompts, tool/function definitions or CRM context across calls that aren't back-to-back, this caching option could reduce per-call input costs; it's worth comparing the 2.5x cache-write cost against expected cache hit rates before enabling it broadly.
🧪 Worth Testing
Testing shows whether the longer cache window actually reduces net cost for your traffic pattern (call frequency, prompt size) versus the higher write price.
Sources
- OpenAI API ChangelogOfficialPrimary sourceOriginal article →„You can now choose 12-hour prompt caching in the API. 12-hour prompt caching keeps shared instructions, tools, and context cached longer to benefit from lower cached-input pricing, even when requests “20 Oct 2026, 03:00
- Published by source
- 20 Oct 2026, 03:00
- Found by our system
- 20 Oct 2026, 21:16
- Summary generated
- 20 Oct 2026, 21:17
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy