What is a token in AI? (Explained for kids)
Updated May 8, 2026 · 280 words
A token is a chunk of text — usually a word or part of a word — that an AI sees as one unit. When you type into ChatGPT, your message is broken into tokens before the AI processes it. "Hello world" is two tokens. "Anti-disestablishmentarianism" is many.
How to explain it to a 7-year-old
🧒 "AI doesn''t read whole sentences in one go — it reads them in tiny pieces. Each piece is called a token. Sometimes a token is a whole word, sometimes a part of a word, like ''play'' + ''ing''."
How to explain it to a 14-year-old
🎒 "Tokens are how language models digest text. The model can''t process raw characters — it splits text into a vocabulary of ~50–100k tokens. Each token gets converted to a number (an embedding) before going into the network. ''Tokens per dollar'' is how AI companies bill API access."
Why tokens matter
- 💰 Pricing — most AI APIs charge per token (~$0.01 per 1,000 tokens for GPT-4o)
- 📏 Limits — models have a "context window" measured in tokens (e.g., GPT-4o handles 128k tokens at once)
- 🌍 Languages — English is token-efficient; Hindi, Japanese, and code take more tokens
Where this comes up in Chippu
Band D (d2-2) introduces tokens for older kids studying how LLMs work.
Related terms
- Large language model
- Embedding — the number a token becomes
- Prompt