Prompt Caching and Context Caching

Your system prompt is 4,000 tokens. Your RAG context is 20,000 tokens. You send both with every request. You also pay for both — every time. Prompt caching lets the provider keep that prefix warm on their side and bill you 10% of the norma...