Glossary / LLM APIs and AI
Glossary · LLM APIs and AI · checked Oct 24, 2026
Prompt caching (cache read)
Short answerPrompt caching lets a provider reuse the processed start of a prompt you send again (a system prompt, a document, an agent's history). The repeated part is billed at a much lower "cache read" price.
Why it matters
For chatbots with a long fixed system prompt and for coding agents that resend the repo context every step, caching is often the biggest single cost cut.
Our number
56 listed models have a cache-read price, a median 10% of the normal input price. On our coding-agent step (40,000 tokens in, 30,000 of them repeated), caching cuts the cost a median 52%.
Go deeper on this site
Where it shows up here
It appears on 78 pages of this site: 75 LLM price pages, 1 Article, 1 Open source, 1 Free tool. A few to start with:
- LLM proxy benchmark: 200 duplicate calls became 1 (and a 30 s bug)
- LLM API cost calculator.
- LLM API prices, and the cheapest model for your job
- chat-latest API pricing
- Free Rust crates you can use today.
Sources
- Anthropic: prompt caching: platform.claude.com/docs/en/build-with-claude/prompt-caching (checked Oct 24, 2026)
- OpenAI: prompt caching: platform.openai.com/docs/guides/prompt-caching (checked Oct 24, 2026)
Related terms
Up next
Go deeperCache prices per modelThe page on this site that does the most with this term.Read it next →Did this answer your question?