Skip to main content

See How an LLM
Thinks

Walk through the 11 stages a large language model uses to process your query — from raw text to generated response. Real tokenization, educational simulations, and interactive visualizations.

Real Tokenization

See exactly how GPT-style models break your text into tokens using real BPE tokenization — not a simulation.

Interactive Visualizations

Explore attention heatmaps, embedding spaces, probability distributions, and more with D3-powered graphics.

Fully Offline

Everything runs in your browser. No GPU, no API keys, no data sent anywhere. Learn at your own pace.