A schematic output projection, followed by softmax and sampling computed from fixed example logits. The diagram values are illustrative.

Pick the next word

Illustrative logits. Temperature changes their sampling probabilities; it does not change model weights.

Final hidden state multiplied by the unembedding matrix to produce logits Three token rows; the last row is highlighted and feeds into a column of six logit scores, one per vocabulary word. final hidden state [3 × 4] × unembedding [4 × vocab] one score per word

softmax turns those scores into probabilities:

the cat sat … awaiting prediction