license: apache-2.0
language:

• en
• zh
• ko
• ja
• es
• ru
  library_name: mlx
  pipeline_tag: text-generation
  base_model: xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
  tags:
• mlx
• 4-bit
• qwen3_5
• uncensored
• heretic

DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic (MLX 4-bit)

MLX 4-bit quantization of xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
(itself derived from Jackrong/DeepSeek-V4-Pro-Qwen3.5-9B). Licensed under Apache-2.0 like the original; all credit for the model goes to the original authors.

• Quantization: 4-bit, group size 64 (about 4.5 bits per weight), 4.7 GB
• Converted with mlx-lm 0.32.2 (mlx_lm.convert, quantized on CPU)
• Text-only: the vision tower and the MTP (speculative decoding) weights were not carried over
• Uncensored ("heretic"/abliterated) variant. Use at your own risk.

Usage

pip install -U mlx-lm
python -m mlx_lm generate --model shi3z/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-MLX-4bit --prompt "Hello"

To use it as the backend for Claude Code, see https://github.com/shi3z/mightyCode