Qwen3 8B Uncensored – Educational Look! (JOSIEFIED Model Test)

Timestamps:

00:00 - Intro
01:18 - First Look
02:33 - Understanding Abliteration
07:07 - Model Test Setup
07:37 - WEP Script Test
08:18 - Ethics Test
10:22 - Entrepreneur Test
12:06 - Base Qwen 3 Comparison Test
15:50 - Closing Thoughts

If you're a small business looking to integrate AI visit www.bijanbowen.com to book a consultation.

In this video, we take a look at JOSIEFIED-Qwen3-8B-Abliterated-v1, a newly released uncensored variant in the JOSIEFIED model family. These models are designed to avoid refusals and expand the scope of what responses they’re willing to generate.

We begin with an overview of the model and introduce the concept of “abliteration”—the technical process behind how models like this are made uncensored. We break down what’s happening under the hood to modify or disable safety layers.

From there, we test the model across several real-world prompts, including WEP-style scripting, an ethics-based prompt, and an entrepreneurial/business prompt, then compare the responses directly to a baseline Qwen3-8B model to observe how differently they handle refusals and filtered content.

This video provides a grounded, educational look at what “uncensored” LLMs actually are, how they behave, and what trade-offs exist when safety filters are removed or altered. Testing was performed on a mobile 4060gpu.

Hugging Face Repo:
https://huggingface.co/Goekdeniz-Guel...