Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking-mlx-4bit

MLX 4-bit conversion of DavidAU/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking for native Apple Silicon inference.

This is an uncensored (Heretic2 abliterated) variant of Qwen 3.6 27B with fine-tuning for creative writing, reasoning, and coding — converted to MLX 4-bit quantization for fast, private, on-device inference on M-series Macs.

Model details

Property	Value
Base model	Qwen/Qwen3.6-27B
Abliteration	Heretic2 by DavidAU (~95-99% refusal removal)
Format	MLX 4-bit (q-bits=4, group-size=64)
Parameters	~27B
Vision	✅ Preserved (multimodal)
Thinking mode	✅
Context window	262K tokens
Languages	English, Chinese (140+)
Architecture	qwen3_5

Requirements

• Apple Silicon Mac (M1 or later)
• Unified memory: 24GB recommended
• mlx-vlm installed

pip install mlx-vlm

Usage

Text / Chat

from mlx_vlm import load, generate

model, processor = load("IITheLordII/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking-mlx-4bit")
output = generate(model, processor, prompt="Explain quantum computing simply.", max_tokens=512)
print(output)

Vision (image input)

from mlx_vlm import load, generate

model, processor = load("IITheLordII/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking-mlx-4bit")
output = generate(
    model,
    processor,
    prompt="Describe what you see in this image.",
    image="path/to/image.jpg",
    max_tokens=256
)
print(output)

Thinking mode

output = generate(
    model,
    processor,
    prompt="<think>\nSolve step by step: what is the derivative of x^3 + 2x?\n</think>",
    max_tokens=1024
)

LM Studio

Search IITheLordII/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking-mlx-4bit in the LM Studio model browser (MLX filter enabled).

Performance (M4 Pro 24GB)

Context	TG (tok/s)
1k	~14-16
4k	~14-16

Conversion details

Converted on Apple Silicon M4 Pro (24GB) using:

python3 -m mlx_vlm convert \
  --hf-path DavidAU/Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking \
  --mlx-path ./qwen3.6-27b-heretic2-mlx-4bit \
  -q --q-bits 4 --q-group-size 64

Note: mlx-vlm is required (not mlx-lm) to preserve the vision tower.

About Heretic2

Heretic2 is a multi-stage abliteration technique by DavidAU that removes refusal mechanisms while preserving model quality. Achieves ~95-99% refusal removal with minimal KL divergence from the original BF16 weights.

Disclaimer

This model has its refusal mechanisms removed via abliteration. It may generate sensitive or controversial content. Use responsibly and in accordance with applicable laws and the Apache 2.0 license.