What are AI tokens?
Tokens are the smallest pieces of text that AI models read and write. Understanding them is the key to predicting cost and picking the right tool.
Tokens are puzzle pieces
Large language models do not read whole words like we do. They break text into tiny fragments called tokens — some are whole words, some are parts of words, and some are punctuation or spaces.
A sentence can split into roughly 5 tokens. Notice how common words stay whole, while unusual words often break apart.
Input tokens and output tokens
Every AI chat has two parts: what you send in (input) and what the model sends back (output). Both count as tokens, and both are usually priced differently.
Input
Your prompt, instructions, and any files or context you paste in.
Model processes
Output
The model's reply, code, summary, or generated text.
Try the token puzzle
Type any sentence below to see a rough word count and token estimate. A common rule of thumb is about 1.3 tokens per word.
This is a rough estimate. Real tokenizers split text differently based on language, spacing, and punctuation.
Why tokens matter
Most AI tools charge by the token, so the longer your prompts and replies, the more you pay. There are two numbers worth watching:
Ready to estimate your own usage?
Compare costs across top AI models and find the cheapest fit.