Okaydvns/ayanokoji
The Okaydvns/ayanokoji model is an uncensored 27 billion parameter language model derived from Qwen/Qwen3.6-27B, created by huihui-ai. It utilizes an abliteration technique to remove refusal behaviors, making it suitable for research into unfiltered language generation. With a 32768 token context length, this model is designed for experimental use where safety filtering has been intentionally reduced.
Loading preview...
Overview
Okaydvns/ayanokoji is a 27 billion parameter language model developed by huihui-ai, based on the Qwen/Qwen3.6-27B architecture. Its primary distinction is the intentional removal of safety filtering and refusal behaviors through an "abliteration" process, a proof-of-concept implementation without TransformerLens. This modification aims to provide an uncensored model for specific research and experimental applications.
Key Characteristics
- Uncensored Output: Safety filtering has been significantly reduced, allowing for potentially sensitive, controversial, or inappropriate content generation.
- Experimental Design: Created as a proof-of-concept to demonstrate refusal removal techniques.
- Qwen3.6-27B Base: Built upon the robust Qwen3.6-27B model, inheriting its foundational capabilities.
- 32K Context Length: Supports a substantial context window for processing longer inputs.
Usage Warnings & Considerations
Due to its uncensored nature, this model comes with significant usage warnings:
- Risk of Sensitive Content: Users must exercise extreme caution as it can generate inappropriate content.
- Not for Public/Production Use: Strongly advised against deployment in public-facing or commercial applications.
- Legal & Ethical Responsibility: Users are solely responsible for ensuring compliance with laws and ethical standards.
- Research Focus: Best suited for controlled research environments and testing.
Good For
- Research into Unfiltered LLM Behavior: Studying the effects of removing safety mechanisms.
- Exploring Content Generation without Refusals: Investigating how models respond to prompts typically flagged by censored versions.
- Controlled Experimental Settings: Environments where outputs can be rigorously monitored and reviewed.