The model catalogue and token prices
See which models Slen can call, what a million tokens of each costs and how many tokens one answer may hold.
2 min read
Every estimate and every reservation is priced from the model catalogue. This page is about the catalogue page of the console: what it lists and how Slen uses its prices.
Where the catalogue is
Open Settings, then Models. The catalogue is the same list in every project and in every organization. Nothing on the page can be changed.
The page opens for whoever has the read permission. Every role has it.
What a row shows
| Column | What it is |
|---|---|
| Model | The model code, such as gpt-4o-mini. A model node names the model by this code. |
| Provider | The provider the model belongs to. Calling the model takes a connection to that provider. |
| Input / 1M | The price of one million input tokens. |
| Output / 1M | The price of one million output tokens. |
| Max output | The largest number of output tokens the model may return in one call. |
How a price is used
A price has a date it takes effect from. A new price does not overwrite the old one. It follows it, and the page shows the price that is in effect now.
A reservation keeps the price it was made with. A later change of a price does not re-price a request that is already reserved.
A model that is not in the list
The list holds the active models that have a price in effect. A model node that names another model cannot make its call.
A model the catalogue does not have, or one that is switched off, is refused with the code
model_not_allowed.A model with no price in effect cannot be priced, and the call is refused with the code
cost_calculation_failed.
Where to go next
Connections — add the provider key a model is called with.
Models — the same catalogue over the API.
Glossary — what a provider, a connection and a pool are.
Error codes — every code of the API, what it means and what to do about it.