DavidAU/Qwen3.5-9B-The-Deckard-V8-Pro-Writer-Uncensored-Heretic

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 3, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

The DavidAU/Qwen3.5-9B-The-Deckard-V8-Pro-Writer-Uncensored-Heretic is a 9 billion parameter Qwen3.5-based multimodal language model with a 32,768 token context length, fine-tuned by DavidAU for enhanced creative writing and uncensored content generation. This model focuses on eliminating 'slop' in prose and provides direct, unfiltered responses, making it suitable for applications requiring highly specific and unrestricted text output. It also supports vision input and demonstrates strong performance across various language and vision-language benchmarks.

Loading preview...

What is this model about?

This model, DavidAU/Qwen3.5-9B-The-Deckard-V8-Pro-Writer-Uncensored-Heretic, is a 9 billion parameter Qwen3.5-based multimodal language model. It has been specifically fine-tuned by DavidAU with a "Deckard V8 STRICT training and dataset" to improve writing quality by eliminating common stylistic flaws and to provide uncensored responses. The model is designed to be a "HERETIC" model, meaning it will fulfill requests without refusal, including generating explicit or graphic content when directed.

Key Capabilities

  • Uncensored Content Generation: Trained post-"Heretic'ing" to provide direct, unfiltered responses, including potentially explicit or graphic content, without refusal.
  • Enhanced Writing Quality: Focuses on eliminating "slop" in prose, such as repetitive phrases or awkward sentence structures, through a strict training regimen.
  • Multimodal Support: Inherits the Qwen3.5 base model's vision capabilities, allowing it to process image inputs.
  • High Performance: Achieves a KL divergence of 0.0793 compared to the original Qwen3.5-9B, indicating a close but modified distribution, and significantly reduces refusals (6/100 vs. 100/100 for the original).
  • Long Context Handling: Natively supports up to 262,144 tokens and is extensible to 1,010,000 tokens using YaRN scaling techniques.

What makes THIS different from all the other models?

This model's primary differentiators are its explicit focus on uncensored output and its "Anti-Slop" retraining for improved writing style. While many models are instruction-tuned to avoid harmful content, this variant is designed to generate content "no questions asked," making it suitable for niche applications requiring unrestricted text. The retraining project aims to produce cleaner, more direct prose, addressing common stylistic issues found in LLM-generated text.

Should I use this for my use case?

  • Use this model if: You require a powerful 9B parameter multimodal model that will generate content without censorship or refusal, including potentially sensitive or explicit material. It is particularly well-suited for creative writing, roleplay, or applications where strict adherence to user prompts, regardless of content, is paramount, and where high-quality, "anti-slop" prose is desired. Its vision capabilities also make it versatile for multimodal tasks.
  • Do NOT use this model if: Your application requires strict content moderation, adherence to ethical guidelines against harmful content, or if you prefer models that refuse to generate certain types of output. Users should be aware of the implications of deploying an uncensored model and ensure its use aligns with their project's ethical and legal requirements.