OpenAI adds 12-hour API prompt caching, with cache writes at 2.5× the input-token price; Tibo says logged-in Codex and Work users now keep caches indefinitely when resuming conversations.

OpenAI’s official interactive prompt-caching diagram shows shared context reused across two requests. Source documentation captured 21 October 2026; illustrative token counts, not a performance benchmark.

Published 21 Oct 2026 · 01:42 UTC

Thibault Sottiaux; OpenAI API changelog ↗ · Source dated 21 Oct 2026

OpenAI adds 12-hour API prompt caching, with cache writes at 2.5× the input-token price; Tibo says logged-in Codex and Work users now keep caches indefinitely when resuming conversations. · ournews.ai