DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking
DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking is a 21 billion parameter Qwen 3.5-based language model, contracted from 27B parameters and fine-tuned on Claude 4.6 Opus High Reasoning datasets. It features variable length reasoning, enhanced performance, and is uncensored, making it suitable for complex, nuanced, and creative text generation. The model supports a 256K context length and excels in scenarios requiring deep thought and character-rich outputs.
Loading preview...
Model Overview
This model, DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking, is a 21 billion parameter (dense) variant of the Qwen 3.5 architecture. It was initially contracted from a 27B Qwen 3.5 base and subsequently fine-tuned using Unsloth on a combination of internal "Deckard/PDK" datasets (focusing on character, intelligence, depth, observation, and point of view) and the Claude 4.6 Opus Distill dataset for improved reasoning and stability. Notably, it is designed to be fully uncensored and unfiltered, offering bold and confident responses.
Key Capabilities
- Advanced Reasoning: Enhanced through Claude 4.6 Opus dataset training, leading to dramatically increased performance.
- Uncensored Output: Provides unfiltered and unaligned responses, suitable for diverse and unrestricted content generation.
- Variable Length Reasoning: Adapts output length based on complexity, delivering shorter responses for simple queries and longer, more detailed ones for complex tasks.
- Character and Depth: Incorporates unique "Deckard/PDK" datasets to imbue responses with distinct character, intelligence, and observational depth.
- Efficient Performance: Operates 33% faster than its 27B base model in terms of tokens per second and requires less memory for quantization.
- Extensive Context: Supports a 256K context window, allowing for processing and generation of very long texts.
Good For
- Creative Writing & Roleplay: Excels in generating dramatic, character-driven narratives and dialogues, as demonstrated by example generations.
- Complex Problem Solving: Its enhanced reasoning capabilities make it suitable for tasks requiring deep thought and nuanced understanding.
- Unrestricted Content Generation: Ideal for applications where censorship or safety alignments are not desired, offering complete freedom in output.
- Code Generation & Image Processing: Tested and confirmed to handle code generation and image processing tasks effectively.