Model parameters like temperature, top-p, and top-k control how the model chooses the next token. They don’t change the model’s knowledge — they change how random or deterministic the output is.
Temperature
How random the next token is. Low (0–0.3) = focused, deterministic. High (0.7–1+) = creative, varied. Try the Temperature simulator.
Top-p (nucleus sampling)
Pick from the smallest set of tokens whose total probability ≥ p. Lower p = more focused; higher = more diverse. Often used with temperature.
Top-k
Only consider the k most likely next tokens. Reduces nonsense from low-probability picks.