Zynerji/Ektome-Qwen3-0.6B-PristinelyUncensored
The Zynerji Ektome-Qwen3-0.6B-PristinelyUncensored is a 0.6 billion parameter language model based on the Qwen3 architecture. It is specifically designed to be uncensored by employing a proprietary 'Ektomē' excision method that removes refusal-specific components while aiming to preserve general helpfulness and knowledge. This model is intended for use cases where an uncensored output is required, with the user accountable for its deployment.
Loading preview...
Overview
Ektome-Qwen3-0.6B-PristinelyUncensored is a 0.6 billion parameter model developed by Zynerji, distinguished by its "uncensored" nature. Unlike standard refusal-removal techniques that can degrade model capability, this model utilizes a proprietary "Ektomē" (excision) method. This technique aims to isolate and remove only refusal-specific components from the pristine model, preserving general helpfulness and knowledge without retraining or distillation.
Key Characteristics
- Uncensored by Design: The model is constructed to not refuse, with the Ektomē method targeting refusal-specific components.
- Capability Preservation Goal: The Ektomē process is designed to maintain the original model's capabilities, with a reported compliance of 0.95 to 1.00 at a capability change of +0.007 versus the pristine Qwen3-0.6B.
- Proprietary Method: The estimator, excision operator, and depth-selection procedure are proprietary to Zynerji.
- No Certification: While point estimates for capability (MMLU-val) and compliance on harmful content are provided, the model is explicitly NOT CERTIFIED as a full n=2800 paired test has not been conducted.
Intended Use Cases
This model is suitable for applications requiring an uncensored language model. Users should be aware that it will not refuse and are accountable for its outputs. It is important to note that the certification process focuses solely on capability retention and does not certify safety, factual accuracy, or fitness for any specific purpose.