LLMTokenBase

Words ↔ tokens, both directions.

Type in either field and the other updates instantly. Handy when you only have a word count from a brief, or a token budget from a model card (32k, 128k, 1M) — not the final paste yet. For OpenAI, tiktoken, Claude, or ChatGPT length on real text, use the live token counter instead.

Ratios are approximate — about 1 token per ¾ of an English word. Actual counts vary by model and language; paste real text into the token counter for a closer estimate.

Common lengths

ContentWords≈ Tokens
A tweet5067
A short email200267
A one-page brief500667
A blog post1,2001,600
A long article3,0004,000

How the ratio works

English text averages roughly 4 characters per token, which works out to about 1 token for every ¾ of a word. This converter uses that single ratio, so it's a planning shortcut — not a substitute for a real tokenizer. Code, punctuation-heavy text, and non-English languages can differ substantially. Read the full methodology →

Good uses for this page

01

You only have a word count from a doc or brief, not the final paste yet.

02

A model card lists 32k, 128k, or 1 million tokens and you want a reading-length intuition.

03

You need a ballpark token budget before drafting, or a word estimate for writers who think in pages.

04

You want a quick ratio check — then verify with the live counter when you have the string.

Related tools

Questions

It's a rough average for English prose (~0.75 words per token). Real tokenizers split text differently; code, CJK, and dense punctuation change the ratio a lot. Treat the result as a ballpark for planning.

Estimates only. Not an official tokenizer from OpenAI, Anthropic, or Google. Verify with your provider before billing decisions.