Mixed-precision (UVMAX) quantization of meituan-longcat/LongCat-2.0, converted with mlx-lm.
Revision note: originally converted from the FP8 release (meituan-longcat/LongCat-2.0-FP8), the current revision is re-converted from the bf16 master checkpoint.
UVMAX is a mixed-precision scheme: bit widths are assigned per tensor class from measured round-trip quantization error, rather than uniformly. All classes use group size 64.
| Tensor class | Bits | Parameters | Size | Share |
|---|---|---|---|---|
| Expert FFNs | 3 | 1.47 T | 598.5 GiB | 88.0% |
| N-gram embedding tables | 3 | 135 B | 55.0 GiB | 8.1% |
| Attention, dense MLPs | 6 | 31.4 B | 22.9 GiB | 3.4% |
Embeddings, lm_head |
6 | 2.8 B | 2.1 GiB | 0.3% |
| DSA indexer, MoE routers | 8 | 0.6 B | 0.6 GiB | 0.1% |
| Norms, correction biases (unquantized) | — | — | 0.9 GiB | 0.1% |
This model requires LongCat-2.0 support from mlx-lm PR #1464, which has not yet been merged. Until it is included in an mlx-lm release, install mlx-lm from the PR branch:
pip install git+https://github.com/ml-explore/mlx-lm.git@refs/pull/1464/head
from mlx_lm import load, generate
model, tokenizer = load("kernelpool/LongCat-2.0-3bit-UVMAX")
prompt = "hello"
if tokenizer.chat_template is not None:
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_dict=False,
)
response = generate(model, tokenizer, prompt=prompt, verbose=True)
model-00067-of-00206.safetensors 5.2 GBmodel-00068-of-00206.safetensors 5.2 GBmodel-00072-of-00206.safetensors 5.2 GBmodel-00073-of-00206.safetensors 5.2 GBmodel-00077-of-00206.safetensors 5.2 GBmodel-00078-of-00206.safetensors 5.2 GBmodel-00082-of-00206.safetensors 5.2 GBmodel-00083-of-00206.safetensors 5.2 GBmodel-00087-of-00206.safetensors 5.2 GBmodel-00088-of-00206.safetensors 5.2 GBmodel-00092-of-00206.safetensors 5.2 GBmodel-00093-of-00206.safetensors 5.2 GBmodel-00097-of-00206.safetensors 5.2 GBmodel-00098-of-00206.safetensors 5.2 GBmodel-00102-of-00206.safetensors 5.2 GBmodel-00103-of-00206.safetensors 5.2 GBmodel-00107-of-00206.safetensors 5.2 GBmodel-00108-of-00206.safetensors 5.2 GBmodel-00112-of-00206.safetensors 5.2 GBmodel-00113-of-00206.safetensors 5.2 GBmodel-00117-of-00206.safetensors 5.2 GBmodel-00118-of-00206.safetensors 5.2 GBmodel-00122-of-00206.safetensors 5.2 GBmodel-00123-of-00206.safetensors 5.2 GBmodel-00127-of-00206.safetensors 5.2 GBmodel-00128-of-00206.safetensors 5.2 GBmodel-00132-of-00206.safetensors 5.2 GBmodel-00133-of-00206.safetensors 5.2 GBmodel-00137-of-00206.safetensors 5.2 GBmodel-00138-of-00206.safetensors 5.2 GBmodel-00142-of-00206.safetensors 5.2 GBmodel-00143-of-00206.safetensors 5.2 GBmodel-00147-of-00206.safetensors 5.2 GBmodel-00148-of-00206.safetensors 5.2 GBmodel-00152-of-00206.safetensors 5.2 GBmodel-00153-of-00206.safetensors 5.2 GBmodel-00157-of-00206.safetensors 5.2 GBmodel-00158-of-00206.safetensors 5.2 GBmodel-00162-of-00206.safetensors 5.2 GBmodel-00163-of-00206.safetensors 5.2 GBmodel-00167-of-00206.safetensors 5.2 GBmodel-00168-of-00206.safetensors 5.2 GBmodel-00172-of-00206.safetensors 5.2 GBmodel-00173-of-00206.safetensors 5.2 GBmodel-00177-of-00206.safetensors 5.2 GBmodel-00178-of-00206.safetensors 5.2 GBmodel-00182-of-00206.safetensors 5.2 GBmodel-00183-of-00206.safetensors 5.2 GBmodel-00187-of-00206.safetensors 5.2 GBmodel-00188-of-00206.safetensors 5.2 GBmodel-00192-of-00206.safetensors 5.2 GBmodel-00193-of-00206.safetensors 5.2 GBmodel-00197-of-00206.safetensors 5.2 GBmodel-00198-of-00206.safetensors 5.2 GBmodel-00202-of-00206.safetensors 5.2 GBmodel-00203-of-00206.safetensors 5.2 GBmodel-00070-of-00206.safetensors 5.2 GBmodel-00075-of-00206.safetensors 5.2 GBmodel-00080-of-00206.safetensors 5.2 GBmodel-00085-of-00206.safetensors 5.2 GBmodel-00090-of-00206.safetensors 5.2 GBmodel-00095-of-00206.safetensors 5.2 GBmodel-00100-of-00206.safetensors 5.2 GBmodel-00105-of-00206.safetensors 5.2 GBmodel-00110-of-00206.safetensors 5.2 GBmodel-00115-of-00206.safetensors 5.2 GBmodel-00120-of-00206.safetensors 5.2 GBmodel-00125-of-00206.safetensors 5.2 GBmodel-00130-of-00206.safetensors 5.2 GBmodel-00135-of-00206.safetensors 5.2 GBmodel-00140-of-00206.safetensors 5.2 GBmodel-00145-of-00206.safetensors 5.2 GBmodel-00150-of-00206.safetensors 5.2 GBmodel-00155-of-00206.safetensors 5.2 GBmodel-00160-of-00206.safetensors 5.2 GBmodel-00165-of-00206.safetensors 5.2 GBmodel-00170-of-00206.safetensors 5.2 GBmodel-00175-of-00206.safetensors 5.2 GBmodel-00180-of-00206.safetensors 5.2 GBmodel-00185-of-00206.safetensors 5.2 GBmodel-00190-of-00206.safetensors 5.2 GBmodel-00195-of-00206.safetensors 5.2 GBmodel-00200-of-00206.safetensors 5.2 GBmodel-00205-of-00206.safetensors 5.2 GBmodel-00017-of-00206.safetensors 5.2 GBmodel-00018-of-00206.safetensors 5.2 GBmodel-00022-of-00206.safetensors 5.2 GBmodel-00023-of-00206.safetensors 5.2 GBmodel-00027-of-00206.safetensors 5.2 GBmodel-00028-of-00206.safetensors 5.2 GBmodel-00032-of-00206.safetensors 5.2 GBmodel-00033-of-00206.safetensors 5.2 GBmodel-00037-of-00206.safetensors 5.2 GBmodel-00038-of-00206.safetensors 5.2 GBmodel-00042-of-00206.safetensors 5.2 GBmodel-00043-of-00206.safetensors 5.2 GBmodel-00047-of-00206.safetensors 5.2 GBmodel-00048-of-00206.safetensors 5.2 GBmodel-00052-of-00206.safetensors 5.2 GBmodel-00053-of-00206.safetensors 5.2 GBmodel-00057-of-00206.safetensors 5.2 GBmodel-00058-of-00206.safetensors 5.2 GBmodel-00062-of-00206.safetensors 5.2 GBmodel-00063-of-00206.safetensors 5.2 GBmodel-00020-of-00206.safetensors 5.2 GBmodel-00025-of-00206.safetensors 5.2 GBmodel-00030-of-00206.safetensors 5.2 GBmodel-00035-of-00206.safetensors 5.2 GBmodel-00040-of-00206.safetensors 5.2 GBmodel-00045-of-00206.safetensors 5.2 GBmodel-00050-of-00206.safetensors 5.2 GBmodel-00055-of-00206.safetensors 5.2 GBmodel-00060-of-00206.safetensors 5.2 GBmodel-00065-of-00206.safetensors 5.2 GBmodel-00001-of-00206.safetensors 4.8 GBmodel-00016-of-00206.safetensors 3.7 GBmodel-00015-of-00206.safetensors 3.7 GBmodel-00014-of-00206.safetensors 3.7 GBmodel-00013-of-00206.safetensors 3.7 GBmodel-00012-of-00206.safetensors 3.7 GBmodel-00011-of-00206.safetensors 3.7 GBmodel-00010-of-00206.safetensors 3.7 GBmodel-00009-of-00206.safetensors 3.7 GBmodel-00008-of-00206.safetensors 3.7 GBmodel-00007-of-00206.safetensors 3.7 GBmodel-00006-of-00206.safetensors 3.7 GBmodel-00005-of-00206.safetensors 3.7 GBmodel-00004-of-00206.safetensors 3.7 GBmodel-00003-of-00206.safetensors 3.7 GBmodel-00002-of-00206.safetensors 3.7 GBmodel-00206-of-00206.safetensors 2.2 GBmodel-00166-of-00206.safetensors 1.1 GBmodel-00106-of-00206.safetensors 1.1 GBmodel-00136-of-00206.safetensors 1.1 GBmodel-00141-of-00206.safetensors 1.1 GBmodel-00126-of-00206.safetensors 1.1 GBmodel-00086-of-00206.safetensors 1.1 GBmodel-00101-of-00206.safetensors 1.1 GBmodel-00151-of-00206.safetensors 1.1 GBmodel-00191-of-00206.safetensors 1.1 GBmodel-00076-of-00206.safetensors 1.1 GBmodel-00091-of-00206.safetensors 1.1 GBmodel-00156-of-00206.safetensors 1.1 GBmodel-00186-of-00206.safetensors 1.1 GBmodel-00161-of-00206.safetensors 1.1 GBmodel-00111-of-00206.safetensors 1.1 GBmodel-00146-of-00206.safetensors 1.1 GBmodel-00201-of-00206.safetensors 1.1 GBmodel-00081-of-00206.safetensors 1.1 GBmodel-00196-of-00206.safetensors 1.1 GBmodel-00071-of-00206.safetensors 1.1 GBmodel-00121-of-00206.safetensors 1.1 GBmodel-00116-of-00206.safetensors 1.1 GBmodel-00176-of-00206.safetensors 1.1 GBmodel-00171-of-00206.safetensors 1.1 GBmodel-00131-of-00206.safetensors 1.1 GBmodel-00096-of-00206.safetensors 1.1 GBmodel-00181-of-00206.safetensors 1.1 GBmodel-00041-of-00206.safetensors 1.1 GBmodel-00051-of-00206.safetensors 1.1 GBmodel-00056-of-00206.safetensors 1.1 GBmodel-00066-of-00206.safetensors 1.1 GBmodel-00046-of-00206.safetensors 1.1 GBmodel-00036-of-00206.safetensors 1.1 GBmodel-00021-of-00206.safetensors 1.1 GBmodel-00031-of-00206.safetensors 1.1 GBmodel-00061-of-00206.safetensors 1.1 GBmodel-00026-of-00206.safetensors 1.1 GBmodel-00069-of-00206.safetensors 805 MBmodel-00074-of-00206.safetensors 805 MBmodel-00079-of-00206.safetensors 805 MBmodel-00084-of-00206.safetensors 805 MBmodel-00089-of-00206.safetensors 805 MBmodel-00094-of-00206.safetensors 805 MBmodel-00099-of-00206.safetensors 805 MBmodel-00104-of-00206.safetensors 805 MBmodel-00109-of-00206.safetensors 805 MBmodel-00114-of-00206.safetensors 805 MBmodel-00119-of-00206.safetensors 805 MBmodel-00124-of-00206.safetensors 805 MBmodel-00129-of-00206.safetensors 805 MBmodel-00134-of-00206.safetensors 805 MBmodel-00139-of-00206.safetensors 805 MBmodel-00144-of-00206.safetensors 805 MBmodel-00149-of-00206.safetensors 805 MBmodel-00154-of-00206.safetensors 805 MBmodel-00159-of-00206.safetensors 805 MBmodel-00164-of-00206.safetensors 805 MBmodel-00169-of-00206.safetensors 805 MBmodel-00174-of-00206.safetensors 805 MBmodel-00179-of-00206.safetensors 805 MBmodel-00184-of-00206.safetensors 805 MBmodel-00189-of-00206.safetensors 805 MBmodel-00194-of-00206.safetensors 805 MBmodel-00199-of-00206.safetensors 805 MBmodel-00204-of-00206.safetensors 805 MBmodel-00019-of-00206.safetensors 805 MBmodel-00024-of-00206.safetensors 805 MBmodel-00029-of-00206.safetensors 805 MBmodel-00034-of-00206.safetensors 805 MBmodel-00039-of-00206.safetensors 805 MBmodel-00044-of-00206.safetensors 805 MBmodel-00049-of-00206.safetensors 805 MBmodel-00054-of-00206.safetensors 805 MBmodel-00059-of-00206.safetensors 805 MBmodel-00064-of-00206.safetensors 805 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 kernelpool/LongCat-2.0-3bit-UVMAX ./model-foldermeituan-longcat/LongCat-2.0