GTWY Navbar

LLM Parameters



#

Mastering LLM Parameters in GTWY: Control How Your AI Thinks, Writes, and Responds

Large Language Models are powerful, but their real strength comes from how you control them.

Without the right configuration, an AI model might generate responses that are too verbose, overly creative, repetitive, or inconsistent. For production systems — whether customer support bots, automation agents, or AI assistants — controlling model behavior is essential.

GTWY solves this by giving you a clear parameter control panel that lets you fine-tune how your AI behaves. Through the Model Settings → Parameters section, you can adjust creativity, response length, diversity, tool usage, and output structure.

Even small adjustments to these parameters can dramatically change how your AI responds.

Below is a simple breakdown of each parameter available in GTWY.


#

1. Creativity Level (Temperature)

The Creativity Level controls how imaginative or predictable the AI’s responses are.

Think of it as the creativity dial for your model.

#

Low Creativity (0.1 – 0.3)

Responses become more logical, focused, and deterministic.

Best used for:

  • Customer support answers

  • Technical explanations

  • Documentation responses

  • Data analysis tasks

#

High Creativity (0.7 – 1.0)

Responses become more expressive and varied.

Best used for:

  • Brainstorming ideas

  • Marketing copy

  • Storytelling

  • Creative content generation

Lower value = accuracy and consistency
Higher value = creativity and variation


#

2. Max Tokens Limit

The Max Tokens Limit controls the maximum length of the AI’s response.

Tokens represent units of text that the model generates. By adjusting this limit, you control how detailed the response can be.

#

Low Token Limit

  • Short answers

  • Quick summaries

  • Concise responses

#

High Token Limit

  • Detailed explanations

  • Long-form responses

  • Complex multi-step outputs

Use this parameter when you want to balance brevity vs depth.


#

3. Probability Cutoff (Top-P)

The Probability Cutoff (Top-P) controls how much variety the model introduces when generating responses.

Instead of choosing from every possible word, the model selects from a subset of likely words based on the probability threshold.

#

Lower Top-P (0.4 – 0.6)

  • Safer responses

  • More predictable outputs

  • Less variation

#

Higher Top-P (0.8 – 1.0)

  • Greater diversity in responses

  • More interesting phrasing

  • Increased creativity

This parameter defines how wide the model’s “idea search space” should be.


#

4. Log Probability

The Log Probability option returns the model’s confidence score for each generated token.

This is primarily useful for:

  • Developers building AI evaluation systems

  • Researchers analyzing model behavior

  • Debugging response quality

If you only need final responses and not analytics, this option can remain disabled.


#

5. Stop (Stop Key)

The Stop parameter defines when the model should stop generating text.

You can specify a keyword, phrase, or symbol that signals the AI to end the response.

Example stop values might include:

  • "stop"

  • "###"

  • "END"

When the model encounters this token, it immediately stops generating further output.

This helps prevent:

  • overly long responses

  • unwanted trailing text

  • incomplete formatting in structured outputs


#

6. Tool Choice

The Tool Choice parameter determines whether the AI should call external tools connected to your GTWY agent.

Within the interface you can select between three options:

#

Auto

The model automatically decides whether to generate a response or call a function/tool.

#

None

The model will not call any tool and will only generate text responses.

#

Required

The model must call one or more tools before producing a response.

This parameter is essential when your AI agent interacts with:

  • APIs

  • databases

  • automation workflows

  • custom functions

It ensures the model uses the right tools at the right time.


#

Quick Parameter Summary

Parameter

What It Controls

Use It For

Creativity Level

AI creativity and randomness

Balance accuracy vs imagination

Max Tokens Limit

Maximum response length

Short answers vs detailed explanations

Probability Cutoff (Top-P)

Diversity of generated responses

Safe vs varied outputs

Log Probability

Token confidence scoring

Debugging and analytics

Stop

Defines where the response ends

Prevent long or unwanted output

Tool Choice

Whether AI calls connected tools

Control integrations and automation


#

Final Thoughts

The Model Parameters panel in GTWY gives you full control over how your AI behaves.

Instead of relying on default behavior, you can fine-tune how the model writes, reasons, and interacts with tools. This flexibility allows you to optimize your AI agents for different scenarios — from structured API workflows to creative content generation.

The key is experimentation.

Small adjustments to parameters like Creativity Level, Top-P, or Max Tokens can significantly change the style, accuracy, and usefulness of the responses your AI generates.

Treat these settings as your AI control dashboard, and you’ll unlock the full potential of GTWY-powered agents.