BAIYUN SI / AI LEXICON
Token The basic unit for processing text
The basic unit for processing information in models is also the basis for measuring usage in many AI services.
In text models, inputs are first split into tokens. A token may be a piece of text, a part of a word, or a symbol; it cannot be fixed and equated with a Chinese character or an English word.
The model reads the input token and then generates the output token. The specific segmentation method, context limit, and output constraints depend on the model and interface; The same passage may count differently in different models.
Input, output, and cache usage may be priced differently. When comparing call costs, you should consider the task results and actual usage, rather than comparing the visible text length of a single request.
Relationship with the Baiyun community
Token supply is now Baiyun's starting point. We want to clarify the model and pricing rules, so that learning, development, and business practice can see our actual consumption. Specific rules are subject to platform announcements.
Further reading
Returning to the journey of light Join the Baiyun SI community