{"version":"2.1.219","anchor":"prompt-caching-doc-rewrites-the-minimum-cacheable-prompt-tab","canonical_anchor":"prompt-caching-doc-rewrites-the-minimum-cacheable-prompt-tab","heading":"Prompt-caching doc rewrites the minimum-cacheable-prompt table and adds cache-preserving escape hatches","tier":"notice","area":"Prompt Caching","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.219\/e\/prompt-caching-doc-rewrites-the-minimum-cacheable-prompt-tab","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.219","markdown":"### Prompt-caching doc rewrites the minimum-cacheable-prompt table and adds cache-preserving escape hatches\n\nCaching docs now list per-model minimums and which request changes keep your cache alive.\n\n**What**\n\nThe per-model minimum table was regrouped and a new table lists which top-level request changes can be made without dropping the cache.\n\n**Details**\n\n- Minimums run from 512 tokens for the newest models up to 4096 for Opus 4.6\/4.5 and Haiku 4.5, with an explicit note that the value is not monotonic across generations.\n- The old Amazon Bedrock override is documented as removed.\n- Cache-preserving forms: tool definitions via `tool_addition`\/`tool_removal` (beta), and system-prompt content via a `role: \"system\"` message with no beta header.\n- Model switch has no escape hatch.\n- Mid-conversation system-message availability widens from one model to four, with a caveat that sources conflict on Sonnet.\n\n**Evidence**\n\n`**The minimum is not monotonic across generations**`\n\n- Area: Prompt Caching\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 0\/5"}