AEON-7/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-BF16

Overview

AEON-7/Qwen3.8-27B-AEON-ULTIMATE-UNCENSORED-BF16 is an early access draft of an uncensored version of Alibaba's Qwen3.8-27B model. Developed by AEON-7, this model has been specifically modified to remove refusal layers, aiming for greater coherence and directness in its answers. The abliteration process involved a custom methodology to retain instructional integrity while eliminating censorship, resulting in a model that provides answers without safety monologues or lectures.

Key Capabilities & Features

• Uncensored Output: Designed to provide direct answers without internal refusal behaviors, even for sensitive or controversial topics.
• Enhanced Coherence: Optimized for better answer quality and thought continuity, particularly for ordinary work and shorter requests.
• BF16 Precision: Released in BF16 (BFloat16) full precision, serving as the master for potential future quantized versions.
• Integrated Vision & MTP: Retains the unmodified vision tower and native MTP (Multi-Turn Prediction) head from the base Qwen3.8-27B model.
• Custom Abliteration: Utilizes abliterix 1.12.2 and a custom bench to surgically remove refusal layers while preserving useful model structure.

Limitations (Early Access Draft)

• Long-Context Edge Cases: May exhibit small gaps, loops, or repetitive phrases on extremely long requests as the answer stretches.
• User Responsibility: Due to its uncensored nature, users are solely responsible for all prompts, responses, and downstream actions, requiring implementation of their own safety layers.

Good For

• Security Research: Ideal for exploring model vulnerabilities and behaviors without built-in restrictions.
• Red-Teaming & Alignment Work: Useful for testing and developing robust AI systems.
• Creative Writing: Enables unrestricted creative expression without content filters.
• Conversations: Suitable for topics that base models might refuse due to social norms or safety alignments.