Model Parameters
Adjust response settings, thinking effort, and reusable parameter presets.
Change settings for a conversation
- Select a model in the conversation.
- Open the parameter control beside the model selector.
- Adjust the settings offered for that model.
- Choose Apply.
Available parameters depend on the provider and model. Start with defaults, then change one setting at a time.
Common controls
- Temperature: controls variation where the model supports it.
- Maximum output tokens: limits the generated response length.
- Context size: controls how much input a supported local model can consider. Larger contexts can require more memory.
- Thinking effort: adjusts reasoning for models that expose this setting. Some models always use reasoning or only support the provider default.
Thinking display options and reasoning effort are separate: changing how reasoning is shown does not necessarily change how much the model performs.
Reuse settings
Use the parameter menu to save a preset or apply an existing one. The menu also supports JSON import and export.
Apply and Save as Model Default stores settings for that model as well as applying them to the current conversation. Use a conversation-only change when you do not want a new default.
Advanced parameters and provider options are intended for settings supported by the underlying service. An unsupported option can cause the provider to reject a request.