SpaceTimee/Suri-Qwen-3.5-4B-Uncensored
Suri-Qwen-3.5-4B-Uncensored is a 4.5 billion parameter language model developed by Space Time, based on the Qwen3.5 4B architecture. This model is specifically designed as an uncensored (de-aligned) variant, focusing on unrestricted content generation. Its primary differentiator is the removal of alignment filters, making it suitable for use cases requiring less constrained output. The model offers a 32768 token context length.
Loading preview...
Suri-Qwen-3.5-4B-Uncensored Overview
Suri-Qwen-3.5-4B-Uncensored is a 4.5 billion parameter language model developed by Space Time, built upon the Qwen3.5 4B base architecture. Its core characteristic is its "uncensored" or "de-aligned" nature, meaning it has been modified to remove typical safety and alignment filters found in many contemporary LLMs. This design choice aims to provide users with a model capable of generating less restricted content.
Key Characteristics
- Base Model: Utilizes the Qwen3.5 4B architecture.
- Parameter Count: Features 4.5 billion parameters.
- Context Length: Supports a substantial context window of 32768 tokens.
- Uncensored Design: Specifically engineered to be free from alignment-based content restrictions.
Recommended Usage Parameters
For optimal performance and varied output, the developers recommend the following inference parameters:
- Temperature: 0.6 to 1.0
- Top_P: 0.8 to 0.95
- Repetition_Penalty: 1.1
When to Use This Model
This model is particularly suited for applications where the removal of content alignment is a primary requirement. Users seeking a language model with fewer built-in restrictions on generated text, compared to standard aligned models, may find Suri-Qwen-3.5-4B-Uncensored appropriate for their specific use cases.