JANG adaptive mixed-precision MLX quantization produced via vmlx / jang-tools.
pip install 'vmlx[jang]'
vmlx serve bearzi/MiMo-V2-Flash-JANG_2S
from jang_tools.loader import load_jang_model
from mlx_lm import generate
model, tokenizer = load_jang_model("bearzi/MiMo-V2-Flash-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))
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.
model-00001-of-00144.safetensors 1.5 GBmodel-00034-of-00144.safetensors 1.3 GBmodel-00067-of-00144.safetensors 1.3 GBmodel-00100-of-00144.safetensors 1.3 GBmodel-00007-of-00144.safetensors 1.2 GBmodel-00010-of-00144.safetensors 1.2 GBmodel-00013-of-00144.safetensors 1.2 GBmodel-00016-of-00144.safetensors 1.2 GBmodel-00019-of-00144.safetensors 1.2 GBmodel-00025-of-00144.safetensors 1.2 GBmodel-00028-of-00144.safetensors 1.2 GBmodel-00037-of-00144.safetensors 1.2 GBmodel-00040-of-00144.safetensors 1.2 GBmodel-00046-of-00144.safetensors 1.2 GBmodel-00049-of-00144.safetensors 1.2 GBmodel-00052-of-00144.safetensors 1.2 GBmodel-00055-of-00144.safetensors 1.2 GBmodel-00058-of-00144.safetensors 1.2 GBmodel-00070-of-00144.safetensors 1.2 GBmodel-00073-of-00144.safetensors 1.2 GBmodel-00076-of-00144.safetensors 1.2 GBmodel-00079-of-00144.safetensors 1.2 GBmodel-00085-of-00144.safetensors 1.2 GBmodel-00088-of-00144.safetensors 1.2 GBmodel-00091-of-00144.safetensors 1.2 GBmodel-00094-of-00144.safetensors 1.2 GBmodel-00106-of-00144.safetensors 1.2 GBmodel-00109-of-00144.safetensors 1.2 GBmodel-00112-of-00144.safetensors 1.2 GBmodel-00115-of-00144.safetensors 1.2 GBmodel-00118-of-00144.safetensors 1.2 GBmodel-00130-of-00144.safetensors 1.2 GBmodel-00133-of-00144.safetensors 1.2 GBmodel-00136-of-00144.safetensors 1.2 GBmodel-00139-of-00144.safetensors 1.2 GBmodel-00004-of-00144.safetensors 1.2 GBmodel-00022-of-00144.safetensors 1.2 GBmodel-00043-of-00144.safetensors 1.2 GBmodel-00061-of-00144.safetensors 1.2 GBmodel-00082-of-00144.safetensors 1.2 GBmodel-00103-of-00144.safetensors 1.2 GBmodel-00121-of-00144.safetensors 1.2 GBmodel-00127-of-00144.safetensors 1.2 GBmodel-00031-of-00144.safetensors 1.1 GBmodel-00064-of-00144.safetensors 1.1 GBmodel-00097-of-00144.safetensors 1.1 GBmodel-00124-of-00144.safetensors 1.1 GBmodel-00003-of-00144.safetensors 872 MBmodel-00006-of-00144.safetensors 872 MBmodel-00009-of-00144.safetensors 872 MBmodel-00012-of-00144.safetensors 872 MBmodel-00015-of-00144.safetensors 872 MBmodel-00018-of-00144.safetensors 872 MBmodel-00021-of-00144.safetensors 872 MBmodel-00024-of-00144.safetensors 872 MBmodel-00027-of-00144.safetensors 872 MBmodel-00030-of-00144.safetensors 872 MBmodel-00036-of-00144.safetensors 872 MBmodel-00039-of-00144.safetensors 872 MBmodel-00042-of-00144.safetensors 872 MBmodel-00045-of-00144.safetensors 872 MBmodel-00048-of-00144.safetensors 872 MBmodel-00051-of-00144.safetensors 872 MBmodel-00054-of-00144.safetensors 872 MBmodel-00057-of-00144.safetensors 872 MBmodel-00060-of-00144.safetensors 872 MBmodel-00063-of-00144.safetensors 872 MBmodel-00069-of-00144.safetensors 872 MBmodel-00072-of-00144.safetensors 872 MBmodel-00075-of-00144.safetensors 872 MBmodel-00078-of-00144.safetensors 872 MBmodel-00081-of-00144.safetensors 872 MBmodel-00084-of-00144.safetensors 872 MBmodel-00087-of-00144.safetensors 872 MBmodel-00090-of-00144.safetensors 872 MBmodel-00093-of-00144.safetensors 872 MBmodel-00096-of-00144.safetensors 872 MBmodel-00102-of-00144.safetensors 872 MBmodel-00105-of-00144.safetensors 872 MBmodel-00108-of-00144.safetensors 872 MBmodel-00111-of-00144.safetensors 872 MBmodel-00114-of-00144.safetensors 872 MBmodel-00117-of-00144.safetensors 872 MBmodel-00120-of-00144.safetensors 872 MBmodel-00123-of-00144.safetensors 872 MBmodel-00033-of-00144.safetensors 872 MBmodel-00066-of-00144.safetensors 872 MBmodel-00099-of-00144.safetensors 872 MBmodel-00126-of-00144.safetensors 872 MBmodel-00129-of-00144.safetensors 872 MBmodel-00132-of-00144.safetensors 872 MBmodel-00135-of-00144.safetensors 872 MBmodel-00138-of-00144.safetensors 872 MBmodel-00141-of-00144.safetensors 872 MBmodel-00142-of-00144.safetensors 820 MBmodel-00002-of-00144.safetensors 604 MBmodel-00005-of-00144.safetensors 604 MBmodel-00008-of-00144.safetensors 604 MBmodel-00011-of-00144.safetensors 604 MBmodel-00014-of-00144.safetensors 604 MBmodel-00017-of-00144.safetensors 604 MBmodel-00020-of-00144.safetensors 604 MBmodel-00023-of-00144.safetensors 604 MBmodel-00026-of-00144.safetensors 604 MBmodel-00029-of-00144.safetensors 604 MBmodel-00035-of-00144.safetensors 604 MBmodel-00038-of-00144.safetensors 604 MBmodel-00041-of-00144.safetensors 604 MBmodel-00044-of-00144.safetensors 604 MBmodel-00047-of-00144.safetensors 604 MBmodel-00050-of-00144.safetensors 604 MBmodel-00053-of-00144.safetensors 604 MBmodel-00056-of-00144.safetensors 604 MBmodel-00059-of-00144.safetensors 604 MBmodel-00062-of-00144.safetensors 604 MBmodel-00068-of-00144.safetensors 604 MBmodel-00071-of-00144.safetensors 604 MBmodel-00074-of-00144.safetensors 604 MBmodel-00077-of-00144.safetensors 604 MBmodel-00080-of-00144.safetensors 604 MBmodel-00083-of-00144.safetensors 604 MBmodel-00086-of-00144.safetensors 604 MBmodel-00089-of-00144.safetensors 604 MBmodel-00092-of-00144.safetensors 604 MBmodel-00095-of-00144.safetensors 604 MBmodel-00101-of-00144.safetensors 604 MBmodel-00104-of-00144.safetensors 604 MBmodel-00107-of-00144.safetensors 604 MBmodel-00110-of-00144.safetensors 604 MBmodel-00113-of-00144.safetensors 604 MBmodel-00116-of-00144.safetensors 604 MBmodel-00119-of-00144.safetensors 604 MBmodel-00122-of-00144.safetensors 604 MBmodel-00032-of-00144.safetensors 604 MBmodel-00065-of-00144.safetensors 604 MBmodel-00098-of-00144.safetensors 604 MBmodel-00125-of-00144.safetensors 604 MBmodel-00128-of-00144.safetensors 604 MBmodel-00131-of-00144.safetensors 604 MBmodel-00134-of-00144.safetensors 604 MBmodel-00137-of-00144.safetensors 604 MBmodel-00140-of-00144.safetensors 604 MBmodel-00143-of-00144.safetensors 543 MBmodel-00144-of-00144.safetensors 278 MBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo bearzi/MiMo-V2-Flash-JANG_2S ./model-folderXiaomiMiMo/MiMo-V2-Flash