FORCE_PROMPT_CACHING_5M

Description

Environment variable shipped in v2.1.108 that forces 5-minute TTL — the cost-correct choice for Max subscribers and workflows where 5m is cheaper than 1h (1h write premium is 100%, 5m is 25%). Pairs with ENABLE_PROMPT_CACHING_1H as the complementary opt-in surface.

Key claims

Relations

Sources

src-20260419-392eceac3ff7