Qwen3.5-27B-JANG_2S / README.md
bearzi's picture
Upload Qwen3.5-27B-JANG_2S
f807f4c verified
metadata
base_model: Qwen/Qwen3.5-27B
library_name: mlx
pipeline_tag: text-generation
license: apache-2.0
tags:
  - mlx
  - jang
  - jang-quantized
  - JANG_2S
  - mixed-precision
  - apple-silicon

Qwen3.5-27B-JANG_2S

JANG adaptive mixed-precision MLX quantization produced via vmlx / jang-tools.

  • Quantization: 2.84b avg, profile JANG_2S, method mse-all, calibration activations
  • Profile: JANG_2S
  • Format: JANG v2 MLX safetensors
  • Compatible with: vmlx, MLX Studio, oMLX (with JANG patch)

Usage

vmlx (recommended)

pip install 'vmlx[jang]'
vmlx serve bearzi/Qwen3.5-27B-JANG_2S

Python

from jang_tools.loader import load_jang_model
from mlx_lm import generate

model, tokenizer = load_jang_model("bearzi/Qwen3.5-27B-JANG_2S")
messages = [{"role": "user", "content": "Hello"}]
prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
print(generate(model, tokenizer, prompt=prompt, max_tokens=512, verbose=True))

About JANG

JANG (Jang Adaptive N-bit Grading) assigns different bit widths to different layer types — attention layers get more bits, MLP/expert layers compress harder. This preserves model coherence at aggressive compression levels where uniform quantization breaks down.

See JANG documentation and scores at jangq.ai.

Comparative benchmarks and feedback welcome — please open a discussion.