[ GLOSSARY ]
What is prompt caching?
Definition
AI conceptsPrompt caching
Prompt caching is a provider feature that reuses the processed start of a prompt when a later request begins with exactly the same text, at a discount.
Why it matters for schools
Caching helps, but less than often claimed. Only cached input gets the discount; output and reasoning tokens are never cached. In our own cost model, caching cut the total bill 27-40%, not 90%. Any change to the start of a prompt breaks the match: as OpenAI puts it, "Cache reuse requires the entire rendered prefix to match."
[ FAQ ]
Questions and answers
Does prompt caching cut AI costs by 90%?
No. In our cost model, total savings were 27-40%, because output is never cached.
What breaks a prompt cache?
Any change to the start of the prompt, such as different instructions or tool definitions.
Does caching change the AI's answers?
Providers say caching does not change output; it changes cost and speed.
[ SOURCES ]
Sources
Facts checked Oct 18, 2026.
- 1OpenAI, prompt caching guideRead Oct 18, 2026 · developers.openai.com
- 2Anthropic, prompt cachingRead Oct 18, 2026 · platform.claude.com
- 3OpenRouter, prompt cachingRead Oct 18, 2026 · openrouter.ai
[ AI FOR SCHOOLS ]
The AI your school can approve.