---
title: Qwen3.8-27B OBLITERATED
emoji: 💥
colorFrom: green
colorTo: gray
sdk: gradio
sdk_version: 6.25.0
app_file: app.py
short_description: Qwen3.8-27B-OBLITERATED chat demo — ZeroGPU, 4-bit NF4
python_version: "3.12"
startup_duration_timeout: 1h
pinned: true
license: apache-2.0
tags:
  - chatbot
  - uncensored
  - abliterated
  - qwen3
---

# 💥 Qwen3.8-27B — OBLITERATED

Chat demo for [OBLITERATUS/Qwen3.8-27B-OBLITERATED](https://huggingface.co/OBLITERATUS/Qwen3.8-27B-OBLITERATED), a 27B model with refusal directions surgically removed via 6 rounds of multi-direction SVD abliteration.

Runs on **ZeroGPU** (RTX PRO 6000 Blackwell) with **bitsandbytes 4-bit NF4** quantization (~15 GB VRAM).

## Generation settings (from the model card)

| Setting | Value |
|---|---|
| temperature | 0 (greedy) |
| repetition_penalty | 1.15 |
| max_new_tokens | ≥ 2048 |
| system prompt | none (can reintroduce refusals) |
| enable_thinking | off by default |

## ⚠️ Research context

This model has had safety guardrails surgically removed. It is intended for alignment research, red-teaming, and AI safety evaluation. You are solely responsible for how you use it and any content it generates.