12-hour prompt caching
OpenAI added a 12-hour prompt caching option to the Responses API, keeping shared instructions, tools and context cached longer at cached-input pricing.
Developers can now choose a 12-hour cache so shared instructions, tools and context stay cached even when requests are hours apart, and still get lower cached-input pricing. 12-hour cache writes cost 2.5x the input token price. An earlier extended prompt caching option kept prefixes active for up to 24 hours.
- Date
- Tuesday 20 October 2026
- Lab
- OpenAI
- Kind
- feature
- Access
- closed API
- Price
- 12-hour cache writes cost 2.5x the input token price
Figures
| Measure | Value | Measured by |
|---|---|---|
| 12-hour cache write cost | 2.5x input token price cache writes for the 12-hour option | company |
The changelog entry links to the prompt caching guide, which was not read, so cached-read prices for this option are not confirmed.
Sources
This record was partly confirmed: some claims could not be checked on 20 October 2026. How we check