One piece at a time.

Where does the next token come from?

TRY THISChange the context or temperature, then choose the next token.

TEACHING SIMULATION0 / 3 TOKENS

The model’s output scores can be normalized into a distribution. Here the scores are authored, and the probabilities are computed.

Start with

The sky is

“blue”64.5%

“cloudy”23.7%

“grey”11.8%

SelectionTemperature · 1.00

Higher temperature spreads probability more evenly. Probability is not factual confidence.

Teaching simulation

What this helps you see

A model assigns probabilities to possible next tokens using the text so far. A token may be a word, part of a word, or punctuation. The chosen token joins the context for the next step.

Keep the limits in view.

The scores below are authored for a small toy vocabulary. They are not probabilities or activations measured from a commercial model.

Follow the sources

Vaswani et al.: Attention Is All You Need OpenAI: Better language models and their implications

Ask about what you’re seeing

Authored learning guide · no live AI generation or web search

NEXT QUESTION

What if the next output requests a tool?