AI Token Counter
Paste a prompt or document — estimated tokens and how much of each context window it fills. (An estimate: ±15% on normal prose.)
| Context window | 8K | 32K | 128K | 200K | 1M |
|---|---|---|---|---|---|
| Filled | 0% | 0% | 0% | 0% | 0% |
Estimated in your browser — prompts never leave this page.
Common questions
What is a token?
The unit LLMs read text in — roughly a word fragment. "Tokenization" splits text so that common words are one token and rarer words several; in English it averages out near 4 characters or ¾ of a word per token.
How accurate is this estimate?
Within about ±15% for normal English prose — it blends the characters÷4 and words×4/3 heuristics. Code, dense punctuation, and non-English text tokenize less predictably. For billing-grade counts, use the provider's own tokenizer; for "will my prompt fit," this is plenty.
Why do tokens matter?
Two reasons: context windows (a model that takes 128K tokens simply can't read more) and cost (API pricing is per token, in and out). Knowing your document is ~40K tokens tells you instantly which models can handle it and roughly what it costs.