Short answerPrompt caching reuses an already-processed repeated start of a prompt, so cached input tokens cost less. Anthropic lists cache reads at a fraction of the base input price; OpenAI lists separate cached-input rates.
The idea
A bot sends the same long instructions with every message. Caching lets the provider reuse the already-processed start of the prompt, which can reduce cost and delay.
- OpenAI: cached input tokens are billed at a lower rate; the pricing page lists a separate cached-input price (for example $0.50 versus $5.00 per million tokens for one model).
- Anthropic: a 5-minute cache write costs 1.25 times the base input price, a 1-hour write 2 times, and a cache read about 0.1 times for most models (lower for some).
How to benefit
- Keep the stable part (instructions) at the start of the prompt and the changing part (question, retrieved passages) at the end.
- Avoid inserting changing values such as timestamps into the instructions.
- Check your platform supports it; most hosts do not expose it.
Related: tokens, reducing cost.
Check your GPT before you move it
Paste your instructions, file names and Actions schema into the free checker. It runs in your browser and builds a portable bundle. Nothing is uploaded.
Open the free checkerRelated guides
Sources
Pages describing other companies' products reflect their public documentation on the date shown and can go out of date.