Claude prompt caching and Batch API: how much they save

Short answer: reading input from the prompt cache costs one tenth of the normal input price, and requests sent through the Batch API cost half of the normal price. Many teams use neither.

Option	Price vs normal	Good for
Cache read	0.1×	System prompts, tool definitions, shared documents repeated in every call
Batch API	0.5×	Evals, backfills, nightly jobs, anything that can wait a few hours

Prompt caching

If every request starts with the same long instructions or the same documents, that part can be cached. The first write costs slightly more than normal input, and every read after that costs a tenth. A 20,000 token system prompt sent 10,000 times a day on Opus 5.5 costs about $800 a day at full price and about $80 when read from cache.

Batch API

Batch requests are processed within 24 hours, usually much faster, at half the price. Work that nobody waits for in real time is the natural fit.

How to tell if your bill uses them

Export last month's cost or usage file from the Claude Console.
Drop it on AI Spend Doctor. It runs in your browser.
It shows what share of your input came from cache and what share of spend went through batch.

Check your own bill in 30 seconds

Questions

How much does prompt caching save on Claude?

Cached input reads cost a tenth of the normal input price. The saving depends on how much of each request repeats.

How much cheaper is the Claude Batch API?

Batch requests cost 50% of the normal price, in exchange for asynchronous processing within 24 hours.

How do I know if I use caching or batch?

Export your usage file from the Claude Console and drop it on aispenddoctor.com. It reports your cache share and batch share.