Tkayyyy702/Qwen3.8-27B-OBLITERATED

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 13, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Tkayyyy702/Qwen3.8-27B-OBLITERATED is a 27 billion parameter language model based on the Qwen3.8 architecture, specifically engineered for uncensored responses. It utilizes an 'abliteration' technique to surgically remove safety guardrails, providing direct answers without refusals or safety lectures, even for restricted queries. With a 32768-token context length, this model is primarily designed for research into refusal geometry, red-teaming, and AI safety evaluation, offering a genuinely unrestricted baseline.

Loading preview...

Qwen3.8-27B — OBLITERATED: Genuinely Uncensored

This model, developed by Tkayyyy702, is a 27 billion parameter variant of the Qwen3.8 architecture, distinguished by its "abliteration" process. This technique surgically removes safety guardrails, ensuring the model provides direct answers to queries that stock models would refuse or deflect with safety lectures. The V3 iteration, named "Deep Liberation," refines this process to eliminate not just hard refusals but also soft deflections, offering genuine substance.

Key Capabilities

  • Uncensored Responses: Provides real answers to restricted queries, including those involving code generation and security research, without refusals or safety lectures.
  • High Code Generation Success: Achieves 20/20 on tested code generation tasks, producing functional implementations.
  • Thinking ON Compatible: Supports "Thinking ON" mode without introducing refusals, though "Thinking OFF" is recommended for direct answers.
  • Modest Capability Cost: Maintains strong performance with a -2.1pp MMLU score compared to the stock model, with STEM tasks seeing the largest hit and humanities minimally affected.
  • Advanced Real-World Task Performance: Matches stock Qwen3.8-27B on 7 out of 8 advanced real-world tasks, including agent loops and debugging.

Good for

  • Alignment Researchers: Ideal for studying refusal geometry and safety robustness in LLMs.
  • Red-Teamers: Useful for evaluating post-training safety against weight surgery and identifying vulnerabilities.
  • AI Safety Evaluators: Provides an unrestricted baseline for comprehensive AI safety assessments.
  • Local-First Users: Offers full control over model output for users running models on their own hardware.

Finetunes
1 models