Cache write
The charge for putting a prefix into the prompt cache. Usually more than the input rate, not less.
Also written: cache writes, cache creation.
Reading from the prompt cache (Paying a reduced rate for a prefix the vendor has already processed, instead of full price for sending it again.) is cheap. Writing to it is not: vendors that publish the figure charge around 1.25 times the input rate for an entry that lives five minutes, and around twice the input rate for one that lives an hour.
That makes the cache a bet. If you read a cached prefix many times before it expires, you win easily. If your context changes between turns — which is exactly what happens when an agent edits files it has already read — you pay the premium repeatedly and can end up spending more than if you had never cached at all.
This is the line almost every pricing comparison on the internet leaves out, and on agentic workloads it is frequently the largest single item on the bill.