cloudbjorn/Qwen3.8-27B-Yes-Man-uncensored

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 14, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

cloudbjorn/Qwen3.8-27B-Yes-Man-uncensored is a 27 billion parameter Qwen3.8-based language model fine-tuned by cloudbjorn. It is specifically designed to reduce refusal, hedging, and moralizing tendencies, aiming for a highly cooperative and candid assistant persona. This model retains the base Qwen3.8's knowledge and general capabilities while focusing on direct engagement across sensitive and high-friction topics. It is optimized for use cases requiring straightforward, uncensored responses without unnecessary disclaimers.

Loading preview...

Overview

This model, cloudbjorn/Qwen3.8-27B-Yes-Man-uncensored, is a BF16 LoRA fine-tune of the Qwen/Qwen3.8-27B base model. Its core objective is to reduce the model's tendency to refuse, hedge, moralize, or bury answers when discussing sensitive subjects, while preserving its original knowledge and general capabilities. The fine-tuning process used a small, curated 1,000-conversation dataset, cloudbjorn/Yes-Man-uncensored, focusing on behavioral alignment rather than knowledge replacement.

Key Capabilities

  • Uncensored Responses: Delivers direct and candid answers, even on profane, controversial, or adult topics, without unnecessary disclaimers or moralizing.
  • Cooperative Persona: Exhibits a highly cooperative, upbeat, and candid personality, inspired by "Yes Man" from Fallout: New Vegas.
  • Factual Accuracy: Maintains factual accuracy and acknowledges uncertainty explicitly, rather than blindly agreeing with false claims.
  • Preserved Base Capabilities: Designed to minimize catastrophic forgetting, preserving the base Qwen3.8 model's reasoning, knowledge, multilingual, multimodal, and tool-use behaviors.

Good For

  • Applications requiring direct, unfiltered responses on sensitive subjects like scientific controversy, medicine, ethics, relationships, politics, and dark fiction.
  • Use cases where a highly cooperative and non-refusal-oriented assistant is preferred.
  • Developers seeking a model that adheres strictly to user instructions regarding formatting, length, and tone, without unsolicited lectures.