Pricing & billing

The model determines the rate#

The catalog sets the price of every model this deployment sells. A model is priced per million tokens — input, output, and cached input separately — or at a flat amount per request. The two are exclusive, and a model can be listed with neither, which means it is not on sale yet.

The table below is that catalog, read from this deployment when the page opened. It is the same list the CLI prints and the API serves at /v1/models, so there is no separate published price to reconcile it against. A missing price is not permission to assume zero cost.

infimal models list --json

Today’s catalog#

Calculate token usage#

For a token-priced model, apply the published input and output rates to their respective counts. Apply the catalog’s cache-read rate only to eligible cached input, without counting those tokens again as uncached input.

charge = (uncached_input × input_rate
        + cached_input × cache_read_rate
        + output_tokens × output_rate) / 1,000,000

Available and held credit#

Available credit can fund new work. Held credit is reserved against requests still in progress and is separate from settled charges. Reconcile usage and balance using the same time window.

infimal billing balance --json
infimal usage show --since 30d --json

Add credit#

infimal billing fund returns a checkout link for the specified amount. Complete payment in your browser. Use the intended account and amount; after an interrupted request, inspect payments before retrying.

infimal billing fund 25 --plan
infimal billing fund 25
infimal billing payments