Ctrl+Enter to process
8
Temperature & Sampling
Temperature controls how "creative" or "random" the model's output is. At low temperatures (close to 0), the model always picks the most likely next word — outputs are predictable and repetitive. At high temperatures (above 1), it spreads the probability more evenly, making surprising or unusual word choices more likely. Temperature of 1.0 means using the model's raw probabilities with no adjustment.
Common misconception: Higher temperature does not make the model "smarter" or more creative in a meaningful sense — it just makes it more willing to pick lower-probability tokens. At very high temperatures the output becomes incoherent. Lower temperature does not mean more accurate; it just means more repetitive and conservative.
Enter a query above to begin.