JoshXT/AGiXT-Qwen3-4B

AGiXT-Qwen3-4B: Core Text Model for Intelligent Agents

This 4 billion parameter text model, fine-tuned by JoshXT from Qwen3-4B-Instruct-2507, is a core component of the AGiXT agent ecosystem. It has been extensively trained on a specialized Agent Interaction Dataset to deeply understand and generate AGiXT's unique operational patterns.

Key Capabilities

• Native AGiXT Command Understanding: Processes and generates AGiXT's XML-based command execution format, including <execute>, <name>, and <param> tags.
• Structured Reasoning: Utilizes <thinking> blocks for internal reasoning before command execution and provides consistent <answer> formatting.
• Intelligent Tool Delegation: Knows when to delegate tasks, such as coding to GitHub Copilot, or using other built-in AGiXT extensions like web_browsing or postgres_database.
• Extension Awareness: Understands the capabilities and proper parameter usage for over 778 AGiXT commands across various extensions.

Good For

• AGiXT Agent Workflows: Ideal for agents requiring precise command generation and structured responses within the AGiXT framework.
• Complex Text-Based Tasks: Handles multi-step reasoning and command execution for tasks that don't require visual input.
• Integration with AGiXT's Routing System: Designed to work seamlessly with the AGiXT-AbilitySelect-270m model for efficient, complexity-based task routing.