s3nh/Qwen3-0.6B-Uncensored
TEXT GENERATIONConcurrency Cost:1Model Size:0.8BQuant:BF16Ctx Length:32kPublished:Mar 22, 2026License:mitArchitecture:Transformer Open Weights Warm
s3nh/Qwen3-0.6B-Uncensored is an 0.8 billion parameter language model based on the Qwen3 architecture, developed by s3nh. This model is specifically designed to be uncensored, aiming for a balance between harmful and harmless outputs. It exhibits a low refusal rate of 3% on the mlabonne dataset, making it suitable for applications requiring direct and unfiltered responses.
Loading preview...
s3nh/Qwen3-0.6B-Uncensored: An Unfiltered Qwen3 Model
This model, developed by s3nh, is an 0.8 billion parameter variant of the Qwen3 architecture, distinguished by its uncensored nature. It is engineered to provide direct responses without the typical content moderation found in many other language models.
Key Capabilities
- Uncensored Output: Designed to generate responses without content filtering, offering a broad range of outputs.
- Balanced Harm/Harmlessness: Aims for a geometric mean between harmful and harmless content, suggesting a nuanced approach to unfiltered generation.
- Low Refusal Rate: Demonstrates a 3% refusal rate based on the mlabonne dataset, indicating a high willingness to respond to prompts.
Good For
- Applications requiring unfiltered and direct language generation.
- Research into uncensored model behavior and safety alignment.
- Use cases where traditional content filters are undesirable or counterproductive.