Model Card for justintime47/qwen3.8-27b-heretic-ara-mlx-mixed_2_6 (MLX)
This is a MLX quantized and formatted version of trohrbaugh/Qwen3.8-27B-heretic-ara, optimized for Apple Silicon architectures.
Model Details
- Base Model:
trohrbaugh/Qwen3.8-27B-heretic-ara - Library Name:
mlx - Pipeline Tag:
image-text-to-text - License:
apache-2.0
Usage with MLX
You can run this model locally using the MLX framework. Ensure you have the mlx-vlm or relevant MLX packages installed.
from mlx_vlm import load, generate
model, processor = load("justintime47/qwen3.8-27b-heretic-ara-mlx-mixed_2_6")
Reproduce it yourself
mlx_vlm.convert --hf-path trohrbaugh/Qwen3.8-27B-heretic-ara \
-q --quant-method awq \
--calibration multimodal \
--quant-predicate mixed_2_6 \
--trust-remote-code \
--mlx-path ./qwen3.8-27b-mlx-mixed_2_6
⚠️ IMPORTANT: This was built using an unmerged branch in
mlx-vlm. To convert Qwen 3.8 27B yourself, check out PR #1899 or wait for it to land upstream.