Quantization format

Q4_K_L

Q4_K_L has a nominal rate of bits per weight. We have no measured files carrying this label yet.

From the file· 565 files measured
Nominal bpw
from the block layout
Measured average
5.203
565 files
Range
3.776–8.749
varies by architecture
File sizes
0.10 GiB+
up to 541.31 GiB

Real files

one per model, most downloaded first
ModelSizeEffective bpwvs nominal
Qwen3.6-35B-A3B21.11 GiB5.043
gemma-4-26B-A4B-it16.03 GiB5.188
gemma-4-12B-it7.36 GiB5.289
Qwen3.5-4B2.95 GiB5.437
Hy3169.99 GiB4.887
gemma-4-31B-it18.57 GiB5.101
Llama-3.2-1B-Instruct0.81 GiB5.640
gemma-4-E2B-it3.85 GiB6.448
Qwen3-8B5.11 GiB5.362
Qwen3.5-0.8B0.60 GiB5.873
Llama-3.1-8B-Instruct4.95 GiB5.291
Ornith-1.0-35B20.27 GiB5.024
Llama-3.2-3B-Instruct1.97 GiB5.266
ThinkingCap-Qwen3.6-27B17.43 GiB5.473
Qwythos-9B-v26.34 GiB5.638
Qwen3.5-122B-A10B72.81 GiB5.000
Qwen3-Coder-Next45.60 GiB4.916
Qwen2.5-32B-Instruct19.03 GiB4.988
Qwen3.5-35B-A3B21.11 GiB5.043
Qwen2.5-Coder-7B-Instruct4.74 GiB5.344
Qwen2.5-7B-Instruct4.74 GiB5.344
Qwen3-0.6B0.56 GiB6.383
Qwen3-VL-2B-Instruct1.10 GiB4.447
Jan-v3-4B-base-instruct2.80 GiB5.449
Qwen2.5-1.5B-Instruct0.97 GiB5.403
From the filewhat these mean

Other quantizations