AI Token

The plain-English definition

A token is the smallest unit of text an AI model handles. It's not exactly a word — it's roughly ¾ of one. The word "thanks" is one token. The word "preposterously" is two or three. Models break your text into tokens before processing, then string tokens together to write back.

The math you actually need to know

• 1 token ≈ 0.75 words
• 100 tokens ≈ 75 words
• 1,000 tokens ≈ 750 words ≈ 1.5 single-spaced pages
• 100,000 tokens ≈ 75,000 words ≈ a short novel

Why tokens matter

Three reasons:

1. Context limits. Models have a maximum context window measured in tokens. Hit it, and earlier text gets forgotten.
2. Pricing. If you use the API, you pay per token. Input tokens (your prompt) and output tokens (the response) often cost different amounts.
3. Rate limits. Free tiers limit your tokens-per-day. Pro tiers let you spend more.

2026 API pricing (per million tokens)

• Claude Sonnet: ~$3 input / ~$15 output
• GPT-5 class: ~$2.50 input / ~$10 output
• Gemini 2.5 Pro: ~$1.25 input / ~$5 output

For everyday chat use (not API), this doesn't matter — you pay a flat $20/month for Pro. Tokens only matter if you build with the API.

How to count tokens

Estimate: divide your word count by 0.75. For exact counts, OpenAI offers a tokenizer at platform.openai.com/tokenizer. Anthropic's tokenizer is similar but differs slightly per model.

Related terms

• Context Window
• Temperature
• System Prompt