What Is Token?

In natural language processing (NLP), a token is the minimum unit into which text is divided, for example, a word, subword, or character. Model inference services process input by converting it into identifiable tokens. After model processing, a token is generated and converted into text or other content carriers required by users. The number of tokens used for model processing, including input and output, is an important measurement unit of model inference service usage.

For details about token billing, see Model Inference Billing Items.