Four rates per model
Input, output, cache read and cache write are quoted per 1M tokens. A 1M token pack is priced at the input rate plus the output rate, which is the simplest full round trip a model can make.
Model marketplace
A pack is added at the price printed on its card. The amount charged at checkout is the sum of the packs in your list plus the digital service fee, and the usage record repeats the same figures.
Every card prints the rates the provider lists for that model, and the pack price is built from them.
Input, output, cache read and cache write are quoted per 1M tokens. A 1M token pack is priced at the input rate plus the output rate, which is the simplest full round trip a model can make.
Grok image models are quoted per request, so their pack is a block of 100 images. The video model with no published rate is listed for reference and cannot be ordered yet.
Every listed model supports prompt caching and tiered routing. The cache lines above show what a cached token costs against the plain input rate.
The pack price column is the amount added to your list when you order one unit.
| Provider | Models | Meter | Pack price range |
|---|