Are token counts exact?

Pasted text is counted with gpt-tokenizer (o200k_base) when it loads. Typed numbers are used as you entered them. Other models split text differently.

What is character ÷ 4?

A rough stand-in used only if there is text and the tokenizer has not returned yet.

Does this send the prompt?

No. Counting stays in the browser.

AI Context Window Calculator

Estimate how much of a model’s context window a prompt and reply would use.

Paste text to count with a tokenizer, or type token numbers. Character ÷ 4 is only a rough stand-in.

Context window

Input text (optional)Expected output (optional)Input tokens

Output tokens

Used

Remaining

128,000

Of window

Typed token counts, or a character-based estimate if you only pasted text and the tokenizer is still loading.

Worked example

Paste a prompt, optionally paste expected output, pick a 128k window, and see used tokens, remaining tokens, and the share of the window.

FAQ

Are token counts exact?: Pasted text is counted with gpt-tokenizer (o200k_base) when it loads. Typed numbers are used as you entered them. Other models split text differently.
What is character ÷ 4?: A rough stand-in used only if there is text and the tokenizer has not returned yet.
Does this send the prompt?: No. Counting stays in the browser.