OpenAI states Codex and ChatGPT Work prompt caches are retained permanently,Responses API cache duration can be extended to 12 hours

AT A GLANCE

The report states that OpenAI permanently retains prompt caches for logged-in Codex and ChatGPT Work users, and provides Responses API with cache durations of up to 12 hours.

Article

OpenAI 28 Day Product Sprint Day16：Codex and Work cache retention is permanent,API extended to 12 hours

Beating Insights AI breaking news,OpenAI Product and Platform Director Thibault Sottiaux announced 28 Day of the Day Product Sprint 16. This update focuses on addressing Agent context-caching issues during long-running operations. For logged-in Codex and ChatGPT Work users,OpenAI announced permanent retention of prompt caches. Even when work resumes after several hours or an old conversation is reopened, cache hits are guaranteed, eliminating the loss of cache-reuse benefits caused by lengthy pauses.API has also been upgraded.OpenAI will extend Responses API optional cache duration from the default 30 minutes to 12 hours. Developers can cache system instructions, tool definitions, and frequently used context for extended periods, allowing Agent to reuse prior computation results when executing tasks again several hours later. The durations can also be used together: fixed instructions are retained for 12 hours, while new conversation content is cached for only 30 minutes.

However,API extending cache duration entails additional costs. 12 The initial write price for an hour cache is 2.5 times that of regular input, twice that of the default 30-minute cache. For subsequent cache hits, input prices for most new models are reduced by 90%，GPT-6.1 Sol is reduced by 95%. Therefore, long-duration caching is better suited to Agent that repeatedly invoke the same context, rather than short tasks executed only once.

Original link https://m.theblockbeats.info/flash/372912

Key points

OpenAI Product and Platform Director Thibault Sottiaux announced 28 Day of the Day Product Sprint 16。
The report states that logged-in Codex and ChatGPT Work users can retain prompt caches permanently.
The report states that Responses API's optional cache duration has been extended from the default 30 minutes to 12 hours.
The article states that developers can combine the durations: fixed instructions are cached for 12 hours, while new conversation content is cached for 30 minutes.
The article states that 12 the initial write price for an hour cache is 2.5 times that of regular input, twice that of the default 30-minute cache.
The article states that, for subsequent cache hits, input prices for most new models are reduced by 90%，GPT-6.1 Sol is reduced by 95%。

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

Caching can be understood as saving previously processed prompts and context for subsequent reuse. The reported update covers Codex, ChatGPT Work, and Responses API；API users can select longer caching, but the initial write cost is higher.

Why it matters to readers

For Agent that resume operation after several hours and repeatedly use the same system instructions or tool definitions, longer caching may reduce the cost of repeated input; however, one-off short tasks may not be suitable for 12-hour caching.

Beginners need to distinguish between cache writes and cache hits: 12 the initial write for an hour cache is more expensive; input prices decline only when it is subsequently hit repeatedly.

Risks and unknowns

The evidence is a media paraphrase and does not include an OpenAI official announcement or original post; the relevant features and pricing have not been verified by a primary source in the evidence provided.
The precise scope of application, privacy rules, and data-retention implications of “permanent prompt cache retention” are not specified.
12 The specific models, regions, and activation requirements supported by hour caching are not specified.
Whether actual costs decline depends on cache-hit rates; the evidence provides no real-world usage data.

OpenAI states Codex and ChatGPT Work prompt caches are retained permanently,Responses API cache duration can be extended to 12 hours

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center