Language Model Settings

AI agents need powerful language models to understand and respond to requests. But different tasks require different levels of capability - a simple chatbot might only need basic comprehension, while a complex analysis tool needs advanced reasoning. That's where language model settings come in - they let you choose and configure the AI engine that powers your agent, ensuring you get the right balance of performance and efficiency for your specific needs.

What are Language Model Settings?

Language model settings control which AI model powers your agent and how it works, directly affecting the quality, speed, and cost of your agent's responses.

Every agent needs a language model to work.

How Language Model Settings Work

When you run an agent, your request is processed according to your chosen model and settings to generate the most appropriate response for your needs.

Configuring Language Model Settings

1. Choose a Model

Available Models

MindPal offers all state-of-the-art AI models. Here's our current model lineup:

How to Choose a Model

Consider these rules of thumbs when selecting a model:

1. Check Model Capabilities

   Make sure the selected model's capabilities match the requirements of the agent's job. Look for these key features in the model tooltip:

   Parameter	Description
   Context window	How much information the model can process at once, including your input and its memory of the conversation
   Maximum output length	The longest response the model can generate in a single turn
   Image processing ability	Whether the model can understand and analyze images you provide
   Tool usage support	Whether the model can use external tools provided in the "Tools" settings of your agent

2. Don't Overkill

   For simple tasks, cost-effective models are often sufficient:

   • Gemini 2.5 Flash (1 credit) - Excellent for most tasks
   • DeepSeek V3/R1 (0.5 credits) - Great for coding and reasoning
   • GPT-5 Mini (3 credits) - Balanced quality and cost

   Reserve premium models (Claude Opus 4.5, o3) for complex reasoning tasks.

3. Consider Model Strengths

   If multiple models meet your requirements, consider their unique strengths:

   Model Family	Key Strengths
   OpenAI (GPT-5, o3)	• Excellent at following instructions • Consistent output quality • Strong at structured data tasks
   Anthropic (Claude)	• Superior coding abilities • Nuanced writing and analysis • Great at technical documentation
   Google (Gemini)	• Massive context windows (1M+ tokens) • Strong reasoning capabilities • Powerful visual understanding
   DeepSeek	• Extremely cost-effective (0.5 credits) • Excellent for coding tasks • Strong reasoning
   Perplexity (Sonar)	• Built-in web search • Real-time information • Research-focused
   Groq (LLaMa, Kimi)	• Very fast responses • Cost-effective • Good for general tasks

2. Set Maximum Output Length

Control how long your agent's responses should be by setting the maximum output tokens.

For your reference, 1,000 tokens is about 750 words.

If you don't set a specific value and use "Auto", the model will adjust the length based on what you're asking it to do.

3. Control Creativity Level

Adjust how creative your agent's responses are by setting the temperature.

The higher the temperature, the more varied and creative the model's responses will be. Good for brainstorming and creative work.

The lower the temperature, the more consistent and predictable the model's responses will be. Good for fact-based tasks.

Pick "Auto" and the model will adjust creativity based on your task.