Interactive LLM Demos
Demo 05

Tokenizer playground

i
Tokenizer playground

A tokenizer splits text into the small pieces a language model can read. A piece may be a word, part of a word, punctuation, or a space.

Edit the sentence, then inspect the model-readable pieces hiding beneath the text.

Characters56
Tokens10
Chars / token5.6

Whitespace often becomes part of the next token. Click any piece to inspect its deterministic teaching ID and bytes.

Selected pieceTokenization
Token ID
i
Token ID

The vocabulary number assigned to a token. The model receives this number rather than the visible text.

1063
UTF-8 bytes
i
UTF-8 bytes

The raw numbers computers use to store text characters. UTF-8 is the common encoding used on the web.

84 111 107 101 110 105 122 97 116 105 111 110
Characters
12