What probably matters to youSection of the release
What
Prompt caching lets the API reuse earlier parts of a conversation to save time and cost. DISABLE_PROMPT_CACHING_HAIKU turns it off for requests to Haiku, the small, fast model Claude Code uses for some background work.
The variable now also turns caching off whenever a request's model resolves to the same model as the default Haiku model. That includes a Haiku model you set with ANTHROPIC_DEFAULT_HAIKU_MODEL. The earlier, narrower check still applies alongside this one.
Why
Setting the variable should now reliably turn caching off for Haiku calls, whatever exact model ID or date suffix they use.
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Confirmed sinceAnthropic's documentation has since written up DISABLE_PROMPT_CACHING_HAIKU, on Model configuration.| `DISABLE_PROMPT_CACHING_HAIKU` | Set to `1` to disable prompt caching for the [default Haiku model](/docs/en/prompt-caching#disable-prompt-caching) |model-configsee the edit
How sure we are
Two sources agreeTwo things we can check say the same as this entry.
Anthropic's documentation agreesAnthropic's documentation has since written up DISABLE_PROMPT_CACHING_HAIKU, on Model configuration.
Anthropic's release notes agreeFixed /model picker showing a hardcoded Haiku version and price when ANTHROPIC_DEFAULT_HAIKU_MODEL pins a different model
