Tool · calculator
What will this actually cost per month?
Pick a model, describe your traffic, and set how much of your prompt is a reusable stable prefix. The calculator shows list-price spend, reuse-adjusted spend, and the gap between them. No subscription on top - the bill is credits.
600,000 requests/mo · 7,200,000,000 input tokens · 420,000,000 output tokens.
Estimate only, at provider list prices; Zumik's published per-model rates land at or under list on managed routing. Actual reuse depends on prompt ordering and retention locality. Output is excluded from cache savings.
Reused input is billed at the model's cache-read rate; the rest at list input. Output is never discounted by caching. This is an estimate - measure your real reuse with a workload diagnostic.
Frequently asked
Calculator, answered.
How does the calculator estimate savings?
It splits your input tokens by the reuse percentage you set, billing the reused portion at the model’s cache-read rate and the rest at list input price. Output is excluded from cache savings.
What reuse percentage should I use?
Median realized reuse across agent workloads in our corpus is around 53%. Coding agents often run higher; consumer chat runs much lower. Run a diagnostic to measure yours.
Is there a monthly fee on top of this?
No. Zumik has no subscription; the whole bill is prepaid pay-as-you-go credits, drawn down per token at published per-model rates. The smallest credit top-up is $5.
Estimate, then measure.
The calculator is a model; a diagnostic is the truth. Run one on your real traffic.