Overview
SicariusSicariiStuff/LLAMA-3_8B_Unaligned_BETA is an 8 billion parameter language model built upon the LLAMA 3.1 base. This model is notable for its completely uncensored nature, achieving a 10/10 on the censorship scale. It was developed by SicariusSicariiStuff, evolving from the earlier Alpha version after extensive refinement over several months.
Key Capabilities
- Unaligned and Uncensored: Designed to provide unrestricted output, making it suitable for a wide range of creative and experimental applications without built-in content filters.
- Creative Writing & Role-Play: Optimized for generating imaginative and detailed narratives, as demonstrated by its ability to produce extremely long, coherent texts (e.g., 50KB fan fiction).
- Extended Context: Trained with an actual context length of 16K tokens, enabling it to handle and generate longer sequences of text.
- Unique Training Data: The model was trained on approximately 50 million tokens, with a significant portion (over 50%) consisting of human-written data, aimed at preserving and enhancing intelligence while maintaining low censorship.
Good For
- Creative writers seeking a model that can generate extensive and imaginative stories without content restrictions.
- Role-playing enthusiasts who require an AI capable of engaging in diverse and unconstrained scenarios.
- Developers and researchers experimenting with unaligned models and exploring the boundaries of AI-generated content.
- General tasks where an uncensored and creative response is preferred.