product · May 21, 2026
Claude Code Implements Prompt Caching with Five-Minute and One-Hour TTL Options
Share the canonical public link.
Claude Code adopted prompt caching to improve speed and cost efficiency for AI applications. The system uses a five-minute TTL by default on API keys or third-party providers and a one-hour TTL on Bedrock, Vertex, Foundry, or Claude Platform that bills cache writes at a higher rate. Claude Code automatically drops to the five-minute TTL when users exceed plan usage limits after May 20 2026 updates.
Below validation threshold — auto-passed without scoring