WithinUsAI/Qwen3-Space.Agent.Claude.Uncensored-4B
WithinUsAI/Qwen3-Space.Agent.Claude.Uncensored-4B is a 4 billion parameter, Qwen3-based agentic reasoning large language model developed by Within Us AI. It is specifically designed to combine strong reasoning capabilities, agentic tool-use behavior, and intentionally reduced refusal rates for uncensored outputs. With a context length of approximately 32K tokens, this model excels in long-context tasks and is suitable for experimentation in agent frameworks and alignment research.
Loading preview...
Model Overview
WithinUsAI/Qwen3-Space.Agent.Claude.Uncensored-4B is a 4 billion parameter, Qwen3-based agentic reasoning LLM developed by Within Us AI. This model is a multi-source merge of Qwen3-derived systems, including reasoning-focused, agent-style, and uncensored variants. Its core design philosophy is to fuse these behaviors, remove typical LLM limits, and maintain a compact size for broad deployment, supporting long-context problem solving and real-world agent pipelines.
Key Capabilities
- Reasoning: Offers step-by-step thinking, multi-hop problem solving, and long-context coherence up to ~32K tokens.
- Agentic Behavior: Supports task decomposition, tool-use compatibility, and structured outputs like JSON.
- Coding: Capable of code generation, debugging, and algorithmic reasoning for SWE-style workflows.
- Uncensored Behavior: Features intentionally reduced refusal rates and more permissive responses, making it suitable for alignment research, safety testing, and edge-case exploration.
Intended Use Cases
- Agent Frameworks: Ideal for tool-calling systems and complex agentic workflows.
- Long-Context Reasoning: Excels in tasks requiring processing and understanding extensive textual information.
- AI Experimentation: Provides a less restricted environment for exploring AI behaviors and capabilities.
- Local Assistants: Suitable for deployment on consumer hardware for personalized, less constrained AI interactions.
- Alignment & Safety Research: Offers a platform for studying model alignment and safety boundaries due to its reduced refusal rates.
Important Considerations
Users should be aware that this model's outputs are less restricted than those from typically aligned models and may generate sensitive or unsafe content. External moderation or guardrails are recommended for production environments.