CezarJedi/Qwen3.5-9B-Abliterated
CezarJedi/Qwen3.5-9B-Abliterated is an uncensored version of the Qwen/Qwen3.5-9B model, developed by CezarJedi. This model has had its safety refusal mechanisms removed using the Heretic tool, allowing it to respond freely without unnecessary refusals while retaining its core intelligence. Based on a hybrid transformer architecture, it is designed for strong instruction-following, complex reasoning, and broad language support, making it suitable for applications requiring unconstrained text generation.
Loading preview...
Overview
CezarJedi/Qwen3.5-9B-Abliterated is an abliterated (uncensored) version of the Qwen/Qwen3.5-9B model. Its safety refusal mechanisms have been surgically removed using the Heretic tool, enabling it to respond freely and fulfill prompts without typical refusals, while preserving its original intelligence and capabilities. This model is currently a work in progress, with development and optimization ongoing.
Key Capabilities
- Uncensored Responses: Designed to provide unrestricted answers by neutralizing refusal vectors.
- Core Intelligence Preserved: Maintains the robust instruction-following, complex reasoning, and broad language support of the base Qwen3.5-9B model.
- Hybrid Transformer Architecture: Utilizes advanced dense attention mechanisms and efficient feed-forward layers.
Good For
- Use cases requiring a language model that does not censor or refuse prompts based on safety guidelines.
- Applications where unconstrained text generation is critical.
- Developers interested in experimenting with uncensored LLM behavior while retaining the base model's performance.