Quantization format

UD_Q5_K_M

UD_Q5_K_M has a nominal rate of bits per weight. We have no measured files carrying this label yet.

From the file· 31 files measured
Nominal bpw
from the block layout
Measured average
7.032
31 files
Range
5.157–15.896
varies by architecture
File sizes
5.92 GiB+
up to 705.94 GiB

Real files

one per model, most downloaded first
ModelSizeEffective bpwvs nominal
Qwen3.6-35B-A3B24.64 GiB5.887
gemma-4-26B-A4B-it19.70 GiB6.374
Qwen-AgentWorld-35B-A3B24.64 GiB6.106
LTX-2.316.79 GiB15.683
GLM-5.2522.31 GiB5.956
Ornith-1.0-35B24.64 GiB6.106
Qwen3.5-122B-A10B87.21 GiB5.989
Qwen3-Coder-Next55.17 GiB5.948
Laguna-S-2.181.83 GiB5.979
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1627.00 GiB7.026
Step-3.7-Flash136.43 GiB5.820
LFM2.5-8B-A1B5.92 GiB6.010
Ornith-1.0-397B275.01 GiB5.954
Qwen3.5-397B-A17B273.50 GiB5.824
North-Mini-Code-1.021.37 GiB6.022
MiniMax-M2.7157.23 GiB5.906
MiniMax-M3296.01 GiB5.954
NVIDIA-Nemotron-3-Super-120B-A12B-BF1699.96 GiB6.947
Mistral-Small-4-119B-260383.04 GiB5.974
Huihui-gemma-4-26B-A4B-it-abliterated19.70 GiB6.374
MiMo-V2.5214.37 GiB5.925
GLM-5.1520.11 GiB5.926
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16397.22 GiB6.087
ERNIE-Image-Turbo6.27 GiB6.708
MiMo-V2.5-Pro705.94 GiB5.926
From the filewhat these mean

Other quantizations