See how AI readsyour words.
Type anything. Watch it get chopped, counted and costed, live.
You write
in words.
It reads
in tokens.
What your text would
cost to run.
Pick a model to see how much of its memory your text fills, and roughly what a single request would cost.
Counts use OpenAI's o200k tokeniser, which is exact for the GPT-5 models here. Other providers tokenise differently, and Anthropic's current models produce around a third more tokens for the same text, so the Claude figures here run low. Pricing and context figures are illustrative and approximate. Provided by the Institute of AI for interest and learning.
Why tokens
matter.
A token is a fragment
Models read text as tokens: short, common chunks of characters. A whole word can be one token, while a rare name or a long number splits into several.
Tokens are not words
English averages roughly one token for every four characters, or about three quarters of a word. Spacing, punctuation and casing all nudge the count.
Tokens are what you pay for
Providers bill per token, and charge more for the tokens a model writes than the ones it reads. Trimming a prompt is the simplest way to cut cost.
Context is a budget
Every model can only hold so many tokens at once. Long documents, chat history and instructions all draw down the same fixed context window.
Membership is free. Accreditation is the standard.
Keep
exploring.
Get AI is a collection of games, experiences, and tools that make AI easier to understand from the Institute of AI.