Token

A token is a piece of text the model processes instead of a whole word. A word may be one token, while a long or rare word may become several pieces; punctuation and whitespace can also be represented as units. The count depends on model and language.

Character count and token count are therefore not the same. In a chat request, the question, retrieved document passages, and generated answer all contribute to the context and can increase usage.

KobiGPT tracks external AI and platform cost with Kobi Kredi. Gemini input and output tokens feed a weighted calculation; output tokens may have a different weight from input tokens.

The result is not a promise of one universal fixed price. Model pricing and calculation functions come from central sources, and an interface estimate need not equal the final usage of a completed turn.

Avoid attaching irrelevant files, choose a focused assistant scope, and keep long documents logically structured. More tokens do not automatically produce a better answer; extra noise can lower retrieval precision.

When comparing costs, ask which model, input and output counts, and rounding rule are used. Check the central application definitions for the Kobi Kredi minimum-turn floor and model-specific pricing.

A useful way to evaluate token is to follow one real question from the source document to the final answer. Record which file was selected, what context reached the model, and what a reviewer would need to verify. This turns a definition into an operational check and makes the result comparable across teams.

The same check should include ownership and change management. Decide who updates the relevant documents, how an outdated result is reported, and which access boundary applies. KobiGPT can provide the assistant and the cited document context, but the organisation still owns the source material, permissions, and the decision made from the answer.