The prompt cache is Anthropic's servers keeping the earlier part of a conversation for a while, so it does not have to be processed again on every turn. A server-side switch can still shorten that cache to 5 minutes for a particular set of models, and subscribers otherwise get 1 hour, as before.
What changed is how Claude Code checks whether the model in use belongs to that set. The rule itself, and its choice between 5 minutes and 1 hour, is the same.
How long the cache lasts affects how much a conversation costs to run and how quickly replies come back, so this matters to subscribers using the models in that set.
tengu_tidy_frost Not enough to sayNothing here resolved what this flag was doing on this version, so nothing here should be read as on or off.
This account: no value returned · anonymous baseline: no value returned · compiled default in v2.1.317: off
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
Read once, for one account on one subscription tier, against v2.1.317. It isn't a statement about your account. What a flag value here can and cannot tell you