DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP

Hugging Face
VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 28, 2026License:apache-2.0Architecture:Transformer0.1K Open Weights Featherless Exclusive Warm

DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP is a 27 billion parameter Qwen3.6-based multi-stage fine-tuned and merged causal language model, developed by DavidAU in collaboration with Nightmedia, TeichAI, armand0e, and trohrbaugh. This model is optimized for general intelligence, instruction following, and problem-solving, notably exceeding 700 ARC-C in both 4-bit and 8-bit quantizations. It features enhanced reasoning, vision capabilities, and is uncensored, making it suitable for a wide range of demanding applications on consumer hardware.

Loading preview...

Overview

DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP is a 27 billion parameter model built on the Qwen3.6 architecture, developed through a collaborative multi-stage fine-tuning and merging process. It is specifically designed to enhance general intelligence, instruction following, and problem-solving abilities on consumer hardware. A key differentiator is its performance, being the first model of its size to exceed 700 ARC-C in both 8-bit and 4-bit quantizations, outperforming the base Qwen3.6-27B in 6 out of 7 benchmarks and matching it on the seventh.

Key Capabilities

  • Superior Intelligence & Problem Solving: Achieves high scores in benchmarks, surpassing base Qwen3.6-27B and Qwen3.6-35B-A3B models.
  • Enhanced Instruction Following: Improved ability to understand and execute complex instructions.
  • Vision Capabilities: Supports image and video input, requiring a separate "mmproj" file for activation.
  • Uncensored Output: "Heretic" fine-tuning by trohrbaugh significantly reduces refusals (4/100 vs. 99/100 for the original model).
  • Optimized Quants: Features NEO IMATRIX GGUF quants with 16-bit output tensors for improved accuracy and multi-token prediction (MTP) for faster inference.

Good For

  • Applications requiring high general intelligence and reasoning.
  • Use cases demanding strong instruction following.
  • Scenarios benefiting from vision-language capabilities.
  • Developers needing an uncensored model for diverse content generation.
  • Deployment on consumer hardware due to optimized quantizations.

Finetunes
1 models