# Prompt Caching

> Prompt caching is a feature offered by some model providers that stores the processed form of a repeated prompt prefix, such as long instructions, a reference document or tool definitions, so later requests that start with the same content can reuse it. Reading cached input is typically billed at a lower rate and processed faster, though some providers charge extra to write to the cache. Caches expire after a short period, and the exact rules vary by provider.

## Why it matters for your business

Many business AI workflows send the same large block of text on every call: a long system prompt, a product catalog, a policy manual or a contract that several questions are asked about. Without caching, you pay full price to process that block each time. With caching, repeated requests cost less and respond faster. To benefit, the stable content must come first in the prompt and the variable content last, and requests need to arrive often enough to keep the cache warm. Check each provider's current pricing and cache rules.

## Example

A contract review tool asks a model 25 separate questions about the same 80-page agreement, one per clause type. The agreement and review instructions sit at the start of each request and are cached after the first question. The remaining 24 requests reuse the cached agreement, so they cost less and return faster than sending the full document fresh each time.

Related terms: https://www.lumeor.studio/glossary/token/, https://www.lumeor.studio/glossary/context-window/, https://www.lumeor.studio/glossary/latency/, https://www.lumeor.studio/glossary/ai-tco-total-cost-of-ownership/, https://www.lumeor.studio/glossary/inference/

---

Source: https://www.lumeor.studio/glossary/prompt-caching/