Skip to Content
FeaturesParameter Settings

Model Parameters

Overview

Model parameters are the key settings that control how AI models generate responses. By fine-tuning these parameters, you can optimize the model’s behavior for your specific use case, whether you need creative writing, factual analysis, or conversational interactions.

Main Dashboard

Core Parameters

Temperature

Controls the randomness and creativity of the model’s responses.

  • Range: 0.0 - 2.0
  • How it works: Lower values make responses more focused and deterministic, while higher values increase creativity and randomness
  • Use cases:
    • 0.0-0.3: Factual responses, code generation, data analysis
    • 0.4-0.7: Balanced responses, general conversation
    • 0.8-2.0: Creative writing, brainstorming, varied outputs

Max Tokens

Sets the maximum length of the model’s response.

  • Purpose: Controls response length and API costs
  • Considerations:
    • Each model has different token limits
    • Longer responses cost more
    • Balance between completeness and efficiency
  • Tips: Start with 1000-2000 tokens for most use cases

Top P (Nucleus Sampling)

Controls the diversity of token selection by considering only the most likely tokens.

  • Range: 0.0 - 1.0
  • Default: 1.0
  • How it works: Lower values focus on more probable tokens, higher values allow more diverse choices
  • Interaction with Temperature: Often used together to fine-tune response quality

Frequency Penalty

Reduces repetition in the model’s responses.

  • Range: -2.0 to 2.0
  • Positive values: Decrease repetition (recommended: 0.1-1.0)
  • Negative values: Increase repetition
  • Best for: Long-form content, creative writing

Presence Penalty

Encourages the model to introduce new topics and concepts.

  • Range: -2.0 to 2.0
  • Positive values: Introduce new topics (recommended: 0.1-1.0)
  • Negative values: Stay focused on existing topics
  • Best for: Exploratory conversations, brainstorming

Reasoning Parameters

Advanced reasoning parameters control how models think through complex problems before generating responses.

OpenAI Models

  • Reasoning Effort: Controls how much computational effort the model puts into reasoning (Low/Medium/High)
  • Reasoning Summary: Determines whether to include reasoning steps in the response

Anthropic Claude

  • Thinking Mode: Enables step-by-step reasoning process (Auto/Manual)
  • Thinking Budget: Sets maximum tokens allocated for internal reasoning

Google AI (Gemini)

  • Enable Thinking: Activates internal reasoning process
  • Thinking Budget: Controls computational resources for reasoning

xAI (Grok)

  • Reasoning Effort: Adjusts reasoning depth and complexity
  • Grok-4: Automatically enables thinking mode (not configurable)

Provider Documentation

For detailed parameter specifications and advanced features:

Last updated on
Multi-model SupportVersioning & History