ChatGPT & Claude Token Counter — Free, Runs in Your Browser

Language models don't read words or characters; they read tokens, chunks of text from a fixed vocabulary. In English a token averages about four characters, so 1,000 tokens is roughly 750 words. Code, numbers, and languages that don't use the Latin alphabet usually take more tokens for the same length of text, which is why a counter beats a word count when you're near a limit.

o200k_base is the vocabulary used by current GPT models, from GPT-4o onwards. cl100k_base is the older one used by GPT-4 and GPT-3.5. The same text gives different counts under each, so pick the one that matches the model you're using. Both are OpenAI's real vocabularies, loaded into this page the first time you choose them.

Anthropic doesn't publish Claude's tokenizer, so no browser tool can count Claude tokens exactly. The estimate here uses the same rule as ChatPeer's in-app context meter: about four characters per token for Latin script, and a denser ratio for other scripts. For an exact figure, Anthropic's API has a token-counting endpoint.

A context window is the most text a model can consider at once: your messages, its replies, and any files or instructions. When a conversation outgrows it, the earliest parts are dropped or summarised, and answers start ignoring things you said at the beginning. The buttons cover common window sizes; enter your model's exact limit in the custom field. Leave headroom, because the model's reply counts against the same window.

Counting happens in JavaScript in this tab. There's no upload and no server to receive what you paste, which matters if the text is a contract, client notes or code. It's the same rule the ChatPeer extension follows: its context meter shows this count live under every Claude and ChatGPT conversation, without the text leaving your browser.