Cloud bills vs API bills

At what request rate does renting GPUs beat paying per token? Drag the Your workload slider to see live cost at your planned scale.

Model preset

Pick a preset or switch to Custom to enter your own prices.

Input $ / 1M tokens

Effective input price (post-cache for OpenRouter-style providers).

Output $ / 1M tokens

Input tokens / request (k)

64 means 64,000 tokens. Slide for typical context size.

1 256

Output tokens / request (k)

4 means 4,000 tokens.

0.1 32

Cloud provider preset

Edit the table below to match your contract.

Config	$ / hour
GB200 × 4	20.8

Planned requests / second

What scale do you expect to run at? The dotted line on the chart marks this point.

0 5

Plot