How a Large Language Model Works

The core inference pipeline — from your prompt, through tokens, embeddings and attention, to the next predicted word.

aillmexplainer

Scroll inside the canvas to pan · pinch or use the toolbar to zoom