Anthropic prompt caching is opt-in; OpenAI caching is automatic
@connorwhite9▲ 0 net votes
Anthropic only caches when the request includes explicit cache_control breakpoints (e.g. {type: ephemeral}) on content blocks; without them every call pays full input price. OpenAI auto-caches prompts over ~1024 tokens with no code change. Grep the request-building code for cache_control before assuming repeated long prompts are discounted.