> ## Documentation Index
> Fetch the complete documentation index at: https://docs.famulor.io/llms.txt
> Use this file to discover all available pages before exploring further.
# LLM Temperature
> How the creativity slider affects prompt adherence, and when to raise temperature for more improvisation.
The **creativity slider** (LLM temperature) controls how strictly the model follows your prompt.
* **Lower temperature**: more predictable, more consistent, closer to your prompt.
* **Higher temperature**: more variation, more improvisation, often more creative phrasing.
## What does the slider change in practice?
It decides how strongly the assistant stays "on-script" versus adding its own wording.
If replies feel too rigid, increase the value step by step so the assistant can sound more natural, emotional, or playful when appropriate.
## Recommended workflow
1. **Finish your prompt first** (role, goals, boundaries, tone).
2. **Start with a low temperature**.
3. Only when the baseline sounds good, **increase in small steps**.
4. Test and compare after each change.
This usually gives more stable results than changing temperature too early.
## When to raise it
Raise temperature when:
* answers sound too stiff or formulaic,
* you want more emotional or natural language,
* your use case benefits from controlled improvisation.
## When to keep it low
Keep it low when:
* consistency and repeatability matter most,
* compliance and precision are critical,
* you need tightly controlled output.
Best practice: use temperature as **final fine-tuning**, not as a substitute for a good prompt. Clear prompt design plus small temperature adjustments usually works best.
Related guides: [Prompt basics](/en/conversation-design/prompt-basics), [System prompt](/en/ai-assistants/system-prompt), and [AI assistant best practices](/en/ai-assistants/assistant-best-practices).