Skip to content

Parameters

Invite-only access

Access to Cleura AI Services is currently invite-only.

Model parameters shape how an LLM composes its responses. We explain two very common parameters, Temperature and Effort, that you can adjust in the Playground.

Temperature

With a low temperature, responses are more predictable, focused, and consistent. With a medium temperature, responses can be more flexible without becoming overly unpredictable. Finally, with a high temperature, responses are more varied, surprising, and creative, but also less coherent and less reliable.

Effort

This controls how much computational effort an LLM puts into working through a problem. While the temperature influences how the LLM presents an answer, the effort influences how much reasoning it performs before it starts composing its response.

When the effort is None, no explicit reasoning-effort level is requested, and the exact behavior depends on the provider. Less effort means faster responses and fewer reasoning tokens. Medium effort means a balanced amount of reasoning and processing time. Finally, High effort means extensive reasoning, potentially better handling of difficult problems, but slower responses and higher token usage.