Token.

What is a token?

A token is the small unit of text an AI language model reads and writes, often a whole word, part of a word or a punctuation mark. Usage, limits and pricing for AI models are measured in tokens.

OpenAI's own rule of thumb for English is that one token is roughly four characters, or about three-quarters of a word. Other languages, code and unusual words break into tokens differently, and each model family counts slightly differently.

Why it matters

If you use AI through an API or a usage-based platform, you pay per token, usually with different rates for what you send in and what comes back. Tokens also set the size of the model's context window, which limits how much text it can consider at once. Long documents, lengthy conversation histories and repeated boilerplate in every request all add tokens, which means higher costs and, eventually, information being left out.

The parts that matter

• Input tokens Everything you send: instructions, documents, conversation history and tool definitions.
• Output tokens Everything the model writes back, usually priced higher than input.
• Reasoning tokens Some models think through a problem before answering. That hidden work is billed as output, so a short answer can cost more than it looks.
• Context window The maximum number of tokens the model can handle in one request, which varies by model.
• Cached tokens Some providers charge less for input they have recently processed, which rewards reusing the same instructions.

What to do about it

If your team uses AI through standard subscriptions, carry on. If you run automations through an API, set a monthly spending limit and alerts in the provider's dashboard, and check usage once a month. Trim instructions repeated in every request, send only the relevant part of long documents, and use a smaller model for routine steps such as sorting or extracting data.