Prompt Caching: The Easiest Win on Repeated Context
Prompt caching is often the highest-return, lowest-effort optimization available: when the same large context repeats across calls, the provider bills the repeated part at a steep discount.
When caching applies
Caching pays off when a big, identical block — a system prompt, knowledge base, code files, few-shot examples — is reused across many requests. The repeated tokens bill at the cached rate, typically a small fraction of normal input.
Structure prompts to hit the cache
Put the stable content at the front and the variable user input at the end, so the cacheable prefix stays identical. Small reorganizations can turn a mostly-uncached prompt into a mostly-cached one.
Calculators for this topic
Frequently asked questions
How much does prompt caching save?+
The repeated portion is billed at the cached rate — often a fraction of the input price — so savings depend on how much of your prompt repeats.
What can be cached?+
Stable, identical context reused across calls: system prompts, knowledge blocks, code and few-shot examples.