AI Token Counter
Estimate how many AI tokens your text uses, and check it against common model context windows.
Words: 0
Characters: 0
Estimated tokens: 0
Estimate only — based on the common ~4-characters-per-token rule of thumb. Each model's real tokenizer may count slightly differently, especially for non-English text or code.
Advertisement space
What This Tool Does
This tool estimates how many AI "tokens" a piece of text will use — the unit AI models actually process text in, and the thing that determines both your context-window limit and, on most APIs, your cost. Paste in a prompt, a document, or anything else you're about to feed to an AI model, and see the estimate update instantly, along with how much of a chosen model's context window it would use. Feeding a long prompt to a newer model like GPT-6 Astra? Check it against that model's context window here first. Not sure which assistant to check against? See our ChatGPT vs Gemini vs Claude comparison for how their context windows and capabilities differ, or our explainer on what changed when Claude Chat and Cowork merged if you're checking a prompt against Claude specifically.
How to Use It
- Paste or type your text into the box.
- Your estimated word count, character count and token count update live as you type.
- Pick a context window size from the dropdown (common presets for today's major models) to see what percentage of that limit your text would use.
Example
Drafting a long system prompt or pasting a document you want an AI model to read in full? Paste it in, check the estimate against a 128K or 200K context window, and know immediately whether it'll fit — before you hit a model's error message for going over the limit. If your prompt itself needs work before you send it, see our guide on writing better ChatGPT prompts.
Frequently Asked Questions
Is this an exact token count?
No — it's an estimate based on the commonly cited rule of thumb that one token is roughly 4 characters of English text (a guideline widely used by AI providers themselves). Each model family's actual tokenizer will give a slightly different exact number, especially for non-English text, code, or heavy punctuation. Use this for a quick sanity check, not for precise API billing.
Why do token counts matter?
Two reasons: most AI APIs charge per token, and every model has a maximum context window — a hard limit on how many tokens of input and output it can handle in one conversation. Going over that limit means your request gets cut off or rejected outright.
Does this work for languages other than English?
The word and character counts work for any language. The token estimate is calibrated for English text specifically — many other languages and scripts use noticeably more tokens per character, so treat the estimate as a rougher approximation for non-English text.
Advertisement space