Quantization format

UD_IQ4_NL

UD_IQ4_NL has a nominal rate of bits per weight. We have no measured files carrying this label yet.

From the file· 29 files measured
Nominal bpw
from the block layout
Measured average
4.012
29 files
Range
3.594–4.734
varies by architecture
File sizes
4.06 GiB+
up to 466.84 GiB

Real files

one per model, most downloaded first
ModelSizeEffective bpwvs nominal
Qwen3.6-35B-A3B16.80 GiB4.014
gemma-4-26B-A4B-it12.68 GiB4.103
DeepSeek-V4-Flash128.43 GiB3.792
Qwen-AgentWorld-35B-A3B16.87 GiB4.181
GLM-5.2347.07 GiB3.958
Ornith-1.0-35B16.87 GiB4.181
Qwen3.5-122B-A10B57.22 GiB3.929
Qwen3-Coder-Next36.54 GiB3.939
Qwen3.5-35B-A3B16.60 GiB3.966
Laguna-S-2.154.71 GiB3.998
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1618.19 GiB4.734
Step-3.7-Flash90.63 GiB3.866
LFM2.5-8B-A1B4.06 GiB4.118
Ornith-1.0-397B182.60 GiB3.953
Qwen3.5-397B-A17B180.45 GiB3.843
North-Mini-Code-1.014.47 GiB4.076
MiniMax-M2.7103.15 GiB3.874
MiniMax-M3197.24 GiB3.967
NVIDIA-Nemotron-3-Super-120B-A12B-BF1660.06 GiB4.173
Mistral-Small-4-119B-260355.13 GiB3.966
Huihui-gemma-4-26B-A4B-it-abliterated12.50 GiB4.044
MiMo-V2.5142.12 GiB3.928
GLM-5.1343.54 GiB3.914
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16273.97 GiB4.199
MiMo-V2.5-Pro466.84 GiB3.919
From the filewhat these mean

Other quantizations