How to use the AI Token Counter
- Paste your prompt or document.
- Choose the model's context window size.
- Read the estimated tokens and remaining space.
What does this tool do?
AI language models read text as tokens — chunks that are often parts of words. In English, a token averages about four characters or three-quarters of a word. Code, numbers and non-English text usually take more tokens.
Each model family uses its own tokenizer, so this tool gives a careful estimate rather than an exact count. For billing-critical counts, use your provider's own token counting.
Why use it?
- Check a document fits in a model's context window.
- Estimate API costs before sending.
- Trim prompts that are too long.
Rule of thumb
1,000 tokens ≈ 750 English words ≈ about 1.5 pages of single-spaced text.
Accuracy and limits
- Estimates are typically within about 10–15% for English prose, less precise for code and other languages.
Privacy
Everything you type is processed in your browser. Nothing you enter is sent to our servers or stored by us. There's no account to create and nothing to install.
Frequently asked questions
Is my prompt sent anywhere?
No. The estimate is calculated in your browser without calling any AI service.
Last reviewed by the M2Toolkit team.