ccharnkij/Gemma-4-E4B-it-Uncensored
Gemma-4-E4B-it-Uncensored by ccharnkij is a 7.9 billion parameter language model fine-tuned from Google's Gemma-4-E4B-it. This model specializes in providing unrestricted and uncensored responses, including adult content, by removing the base model's default safety filters and refusal behaviors. It is designed for open-ended conversations where the base model would typically decline to respond, making it suitable for research and educational purposes requiring unfiltered output.
Loading preview...
Model Overview
ccharnkij/Gemma-4-E4B-it-Uncensored is a fine-tuned version of Google's gemma-4-E4B-it model, specifically modified to remove inherent safety filters and refusal behaviors. This 7.9 billion parameter model, developed by ccharnkij, utilizes LoRA (Low-Rank Adaptation) for fine-tuning and is distributed in the Safetensors format. It supports both text and image modalities.
Key Differentiators
- Uncensored Responses: The primary distinction of this model is its ability to generate content that the base
gemma-4-E4B-itmodel would typically decline, including adult content. - Removed Safety Filters: It bypasses the default safety mechanisms, allowing for more open-ended and unrestricted conversations.
- Reduced Refusal Behaviors: The model is less prone to refusing prompts, even those on sensitive topics, though occasional refusals may still occur, requiring regeneration.
Usage Notes
For optimal performance with sensitive topics, it is recommended to start a fresh chat and avoid mixing sensitive and mundane topics. The model's performance is noted to be better when not using 'thinking' processes. A quantized GGUF version is also available for use with llama.cpp, Ollama, and text-generation-webui.
Intended Use
This model is explicitly intended for research and educational purposes only, providing a tool for exploring unfiltered language model capabilities. It is licensed under the Apache 2.0 License.