Skills
Opus 5 halves the prompt-cache minimum to 512 tokens — short system prompts that never cached now do
The minimum cacheable prompt length on Opus 5 is 512 tokens, down from 1,024 on Opus 4.8, and Anthropic notes prompts too short to cache before "can now create cache entries with no code changes." For high-volume, short-prompt workloads — classifiers, routers, extraction endpoints, small subagents — this is a free cost reduction that requires only setting `cache_control`, since caching remains opt-in and a request without it pays full input price on every call. Worth auditing any service where you previously concluded caching wasn't worth wiring up because the prefix was under 1k tokens.
↳ Follow the thread