All guides
Terminology· 2 min

What are tokens?

Tokens are the chunks of text LLMs actually see — usually 3-4 characters or half a word each.

Key idea

Text is split by a tokenizer (BPE, SentencePiece) before the model touches it. Pricing, context limits and speed are all measured in tokens.

Example

'ChatGPT is amazing!' ≈ 6 tokens. English averages ~0.75 words per token; code and non-English languages are far less efficient.

What to remember

  • Non-English text often costs 2-4× more tokens
  • Emoji and code use many tokens per character
  • Use OpenAI's tiktoken or model-specific tokenizer to estimate cost

Further reading