Observed source change · Anthropic
Claude prompt cache policy and minimums documentation changed
Published documentation; runtime behavior and legal interpretation not established.
Last unchanged check: 2026-10-22T19:50:11Z
First changed check: 2026-10-22T21:55:09Z
Before
On the Claude API, [Claude Platform on AWS](https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws), [Google Cloud](https://platform.claude.com/docs/en/build-with-claude/claude-on-vertex-ai), and [Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry), the minimum cacheable prompt length is: * 1,024 tokens for Claude Opus 4.8, Claude Sonnet 5, Claude Sonnet 4.6, Claude Sonnet 4.5 ([deprecated](https://platform.claude.com/docs/en/about-claude/model-deprecations)), Claude Opus 4.1 ([retired, except on Bedrock and Google Cloud](https://platform.claude.com/docs/en/about-claude/model-deprecations)), Claude Opus 4 ([retired, except on Google Cloud](https://platform.claude.com/docs/en/about-claude/model-deprecations)), and Claude Sonnet 4 ([retired, except on Bedrock and Google Cloud](https://platform.claude.com/docs/en/about-claude/model-deprecations)) * 2,048 tokens for Claude Haiku 3.5 ([retired, except on Google Cloud](https://platform.claude.com/docs/en/about-claude/model-deprecations)) Prompt caching uses [workspace](https://platform.claude.com/docs/en/manage-claude/workspaces)-level isolation. Caches are isolated per workspace, ensuring data separation …
After
On the Claude API, [Claude Platform on AWS](https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws), [Claude Models on Google Cloud (Google managed)](https://platform.claude.com/docs/en/build-with-claude/claude-models-on-google-cloud/google-managed), and [Microsoft Foundry](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry), the minimum cacheable prompt length is: * 1,024 tokens for Claude Opus 4.8, Claude Sonnet 5, Claude Sonnet 4.6, Claude Sonnet 4.5 ([deprecated](https://platform.claude.com/docs/en/about-claude/model-deprecations)), Claude Opus 4.1 ([retired, except on Bedrock and Claude Models on Google Cloud (Google managed)](https://platform.claude.com/docs/en/about-claude/model-deprecations)), Claude Opus 4 ([retired, except on Claude Models on Google Cloud (Google managed)](https://platform.claude.com/docs/en/about-claude/model-deprecations)), and Claude Sonnet 4 ([retired, except on Bedrock and Claude Models on Google Cloud (Google managed)](https://platform.claude.com/docs/en/about-claude/model-deprecations)) * 2,048 tokens for Claude …
Capture provenance and limitations
Raw before SHA-256: 09e35a532ffa5f8b5c462d95100aa4fb84ea9dc2184e67a09cf9673916d00a25
Raw after SHA-256: fab7f318164b6e19fc99076258b1007b2b67b56892a5e766b62f218294f87297
Full source responses are retained in the private evidence archive. Public excerpts show changed wording and may be truncated. The observed interval does not establish the vendor’s exact edit time. Hashes identify retained bytes; they are not an independent timestamp certificate.
Download archived excerpts and capture metadata · Back to the tracker