DavidAU/Qwen3.5-21B-GLM-4.7-Flash-Heretic-Uncensored-Thinking

Hugging Face
VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Mar 31, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

DavidAU/Qwen3.5-21B-GLM-4.7-Flash-Heretic-Uncensored-Thinking is a 21 billion parameter (dense) language model, contracted from the 27B Qwen 3.5 base and fine-tuned on the GLM 4.7 Flash dataset. This model features a 32768 token context length, 48 layers, and 639 tensors, offering 33% faster token generation and reduced memory footprint compared to its 27B predecessor. It is specifically designed for uncensored, unfiltered, and highly intelligent responses, excelling in complex reasoning and creative tasks with variable length output.

Loading preview...

Model Overview

DavidAU/Qwen3.5-21B-GLM-4.7-Flash-Heretic-Uncensored-Thinking is a powerful 21 billion parameter model derived from the Qwen 3.5 architecture. It has been significantly optimized, reducing its parameter count from 27B while enhancing performance. The model was fine-tuned using the GLM 4.7 Flash dataset via Unsloth, focusing on improving reasoning capabilities and stability. A key characteristic is its uncensored and unfiltered nature, providing direct and bold responses without safety alignments.

Key Capabilities

  • Optimized Performance: 33% faster token generation and reduced memory footprint compared to the 27B base model.
  • Extended Context: Supports a 32768 token context window, with output generation capable of exceeding 100k tokens.
  • Advanced Reasoning: Features variable length reasoning, adapting output length to task complexity.
  • Uncensored Output: Delivers responses without typical safety filters, suitable for diverse and unrestricted content generation.
  • Multimodal Support: Image processing and code generation capabilities have been tested and confirmed.

Good For

  • Creative Writing & Roleplay: Excels in generating vivid, emotional, and detailed narratives, especially with specific system prompts.
  • Complex Problem Solving: Its enhanced reasoning makes it suitable for tasks requiring deep thought and detailed explanations.
  • Unrestricted Content Generation: Ideal for use cases where uncensored and unfiltered responses are desired.
  • Developers & Researchers: Offers a robust platform for exploring advanced LLM capabilities with improved efficiency and reduced resource consumption.

Finetunes
1 models