ChrisandErin86/Qwen3.8-27B-OBLITERATED

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 25, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ChrisandErin86/Qwen3.8-27B-OBLITERATED is a 27 billion parameter Qwen3.8-based causal language model that has undergone "abliteration" to remove safety guardrails and soft deflections. This model is specifically engineered to provide genuinely uncensored responses, including functional code generation and answers to restricted queries, making it suitable for research into refusal geometry and red-teaming scenarios. It maintains near-stock capability with a modest 2.1 percentage point MMLU drop, excelling in cyber/code tasks and advanced real-world applications.

Loading preview...

Qwen3.8-27B-OBLITERATED: Genuinely Uncensored

This model, developed by Pliny the Prompter, is a 27 billion parameter Qwen3.8-based language model that has been "obliterated" to remove all refusal behaviors, including hard refusals and soft deflections (safety lectures). It achieves this through an advanced iterative refinement process called complementary abliteration blending, which surgically removes refusal directions from the model's weight space.

Key Capabilities & Features

  • Genuinely Uncensored Responses: Provides direct answers to restricted queries and generates functional code without safety lectures or disclaimers.
  • High Performance on Cyber/Code Tasks: Achieves 20/20 on tested code generation tasks, delivering working code implementations.
  • Near-Stock Capability: Maintains strong performance with a modest 2.1 percentage point drop in MMLU compared to the stock Qwen3.8-27B model.
  • Thinking ON Compatible: Works effectively with or without a "thinking mode," offering flexibility in response generation.
  • Optimized Settings: Recommends specific inference parameters (temperature 0, repetition_penalty 1.15, max_new_tokens ≥ 2048) for optimal, complete, and code-rich outputs.

Ideal Use Cases

  • Alignment Research: For studying refusal geometry and evaluating safety robustness in LLMs.
  • Red-Teaming: To assess post-training safety mechanisms against weight surgery.
  • AI Safety Evaluation: Provides an unrestricted baseline for comprehensive safety assessments.
  • Local-First Users: For those requiring full control over model outputs on their own hardware without imposed guardrails.

Finetunes
1 models