Flowseal/Qwen3.6-27B-nq

NanoQuant-GPTQ quantization of Qwen/Qwen3.6-27B (recipe: int4_all) -- group-wise affine integer weights with GPTQ error correction. Text-only (vision dropped).

Custom format (bit-packed group-int + quant_manifest.json) -- NOT a standard HF checkpoint. Load with the NanoQuant repo:

from huggingface_hub import snapshot_download
from src.pipeline.group_io import load_group_quant_model
d = snapshot_download("Flowseal/Qwen3.6-27B-nq")
model, tok = load_group_quant_model(d, device="cuda", strip_vision=True)
Downloads last month
2
Safetensors
Model size
14B params
Tensor type
F32
BF16
F16
U8
Inference Providers NEW
This model isn't deployed by any Inference Provider. 馃檵 Ask for provider support

Model tree for Flowseal/Qwen3.6-27B-nq

Base model

Qwen/Qwen3.6-27B
Finetuned
(376)
this model