How cost is calculated
Where the cost figures come from, why some models show a cost of zero, and what that means for self-hosted models.
Every model call is recorded when it happens, with its token counts, cost, any error, and the Coworker, knowledge base or system job it was for. The Usage & cost view adds these records up. It never estimates: each figure is a sum of recorded calls.
Where the cost of a call comes from#
When a call finishes, the product prices it from the model's published per-token rates and the tokens the call used. That price is stored with the call. Chat calls and embedding calls are priced the same way.
Token counts and call counts are always recorded.
Models with no known price#
A model whose rates are not known to the product is recorded with a cost of $0.00. This applies to models you serve yourself, such as an Ollama model or another OpenAI-compatible endpoint you host, and to hosted models too new or too obscure to have a published price.
For those models:
- Calls, Input tokens, Output tokens and Errors are still complete.
- Cost understates what you actually spend, because it counts only models with known prices. For a self-hosted model, your real cost is your own infrastructure.
Treat the Total cost tile as the cost of priced models. Use the token columns to compare volume across models that have no price.
Cost reflects the prices at the time of each call. It is not a bill: your provider's invoice is the authority for what you owe.