Concept

Tokenization

Process of splitting text into the discrete units (tokens) that a model processes.

Definition

Tokenization breaks raw text into a model's vocabulary entries — usually subwords via Byte-Pair Encoding or SentencePiece. Token counts determine context-window usage and API billing. The same string can tokenise very differently across providers, which is why exact counts must always be measured.

Common use cases

  • Cost estimation
  • Context budgeting
  • Tokenizer training

Related terms

    Tokenization — AI Glossary | Railwail