Training Data & Temperature
Discover what data goes into building an LLM, and learn how the temperature parameter controls whether outputs are predictable or wildly creative.
Training Data Explorer
Click any category to explore what sources LLMs learn from, why each data type matters, and the potential issues it introduces.
Training Data Growth Over Time
Approximate dataset sizes used to train successive models
Key insight: Training data is the single biggest factor in what an LLM "knows" and how it behaves. Every bias, every gap, and every strength traces back to the data it learned from. Understanding this helps you anticipate where models will excel and where they'll struggle.
Temperature Output Slider
Choose a prompt, then drag the temperature slider to see how the same question produces vastly different outputs from deterministic to experimental.
Key insight: Temperature doesn't make the model "smarter" or "dumber" — it controls randomness. Low temperatures pick the most probable next word every time. High temperatures let less likely words win, producing surprising and sometimes nonsensical results. Matching temperature to your task is one of the most impactful settings you can learn.