GPT-2

The original 124M-parameter GPT-2, running entirely in your browser. It only ever does one thing: predict the next token.

Model loads on your first run.

Output appears here.

Sampling settings

Max new tokens 120

Temperature 0.80

Top-p 0.95

Top-k 50

Repetition penalty 1.10

GPT-2's raw output — the probability it assigns to every possible next token. ␣ marks a leading space.

Show top 10 tokens