CCDV-F

By Anthropic · Question Mode

QuestionQ50

Prompt and Context Engineering

You are configuring a Claude application that will handle a mix of multi-turn conversations and one-off requests. You want to use caching methods to lower token costs where applicable. A teammate also proposes caching the model’s output so the application need not make duplicate Claude calls when similar queries arrive. You would use prompt caching for ...

Explanation

Prompt caching reuses an identical prefix of an API prompt to reduce input-token processing time and cost. The reusable prefix should contain static content that recurs across requests, including system prompts, instructions, shared context, examples, and tool definitions. It does not cache generated model output; application-level response caching is a separate technique.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!