Your provider bills you for what you send. Reduction cuts what they bill for on a call you were already making, and the difference shows up on their invoice rather than ours.
You keep your provider, your account, your model choice and your billing relationship. The only thing that changes is how much of your context they charge you for.
The model still receives what it needs to answer correctly. You are not trading quality for cost, and you are not switching to something cheaper and dumber.
What was sent can be reconstructed exactly. Nothing is summarised away, guessed at, or silently dropped from your context.
You call your provider yourself, under your own credentials. We never hold your provider key, so there is nothing on our side to leak.
Send the context you were going to send anyway.
We return a smaller payload that carries the same meaning.
You call your provider with it, under your own key.
A signed record of what was reduced and what it saved.
Every efficiency claim in this industry is a vendor grading their own work. This one is not. The savings appear as a smaller bill from a company with no stake in making us look good.
We report in dollars, because dollars are the thing you can check. A percentage would be our number. An invoice is theirs.
We are not going to quote you a ratio. A ratio is a number we produce about our own work on a workload we chose, and it tells you nothing about what your traffic will do.
What we will tell you is what your bill did. If it did not move, you do not owe us a share of a saving that did not happen.
Bring your own provider key. Billing stays with your provider. The difference lands on their invoice.