AITRADER/Huihui-Qwen3.5-0.8B-abliterated-mxfp4-MLX

Huihui-Qwen3.5-0.8B-abliterated — MXFP4 MLX

MLX-converted version of huihui-ai/Huihui-Qwen3.5-0.8B-abliterated for Apple Silicon, quantized with MXFP4.

Model Details

Property	Value
Base model	Qwen3.5-0.8B
Type	Vision-Language Model (VLM)
Format	MLX MXFP4 (~5.6 bits/weight)
Size	~0.6 GB
Abliterated	Yes — censorship layers removed

MXFP4 provides maximum compression (~3x vs fp16) while retaining good quality. Best for memory-constrained devices.

Variants

Variant	Size	Quality	Link
fp16	~1.75 GB	Highest	fp16
MXFP8	~0.98 GB	Near-native	mxfp8
MXFP4	~0.6 GB	Good	This repo

Usage

pip install mlx-vlm

# Text generation
python -m mlx_vlm.generate \
  --model AITRADER/Huihui-Qwen3.5-0.8B-abliterated-mxfp4-MLX \
  --prompt "Describe this image in detail" \
  --image <path-or-url>

# Chat UI
python -m mlx_vlm.chat_ui \
  --model AITRADER/Huihui-Qwen3.5-0.8B-abliterated-mxfp4-MLX

Credits

• Original model: huihui-ai
• Base architecture: Qwen
• MLX conversion: AITRADER