# Sampling Parameters

> Settings that influence how predictable or varied a model's generated tokens are.

- Category: Reasoning & Inference
- Stage: Established
- Also called: Temperature, Top-p, Top-k
- Canonical: https://www.joelcsaji.com/ai-dictionary/sampling-parameters

## Plain English

They are knobs for making output more focused or more surprising.

## Example

A classifier uses low temperature for consistency, while a brainstorming tool may allow more variation.

## Why it matters

Sampling settings shape reliability, creativity, reproducibility, and debugging expectations.

## Related

- [Token](https://www.joelcsaji.com/ai-dictionary/token)
- [Prompt](https://www.joelcsaji.com/ai-dictionary/prompt)
- [Evals](https://www.joelcsaji.com/ai-dictionary/evals)

## Sources

- [OpenAI: Advanced Text Generation](https://platform.openai.com/docs/guides/text-generation/parameter-details)