Tools
Token counter
Paste a text and see exactly where the cuts land, token by token, with their ids. The o200k, cl100k, p50k and r50k vocabularies run in your browser and give an exact count for OpenAI models. Anthropic publishes no local tokenizer: for Claude the tool shows an approximation and says so.
Tokenizer
OpenAI
Meta Llama
Everything else
GPT-4o, GPT-4.1, GPT-5.x and the o-series.
A vocabulary downloads once, the first time you pick it, and then stays for the session.
Tokens
0
exact
Characters
323
Words
57
Chars / token
-
323 bytes UTF-8
Breakdown
The vocabulary runs in your browser and your text never leaves it.