Quantization format

Q2_K_S

Q2_K_S has a nominal rate of bits per weight. We have no measured files carrying this label yet.

From the file· 86 files measured
Nominal bpw
from the block layout
Measured average
3.184
86 files
Range
2.701–14.149
varies by architecture
File sizes
0.03 GiB+
up to 74.13 GiB

Real files

one per model, most downloaded first
ModelSizeEffective bpwvs nominal
Qwen-Image-Edit-25096.55 GiB2.752
Codestral-22B-v0.17.12 GiB2.751
DeepSeek-Coder-V2-Lite-Instruct6.01 GiB3.288
Hermes-4-14B5.02 GiB2.920
Qwen-Image-Edit6.55 GiB2.752
Hermes-3-Llama-3.1-8B2.78 GiB2.978
Qwen2-0.5B-Instruct0.31 GiB5.361
Qwen2.5-Coder-32B10.70 GiB2.805
Equinox-31B10.22 GiB2.808
FLUX.1-Krea-dev3.83 GiB2.762
gemma-2-27b-it9.06 GiB2.857
Llama-3.1-70B22.79 GiB2.775
glm-4-9b-chat3.69 GiB3.369
Wan_2.2_ComfyUI_Repackaged8.86 GiB
Huihui-Qwen3-VL-30B-A3B-Instruct-abliterated9.80 GiB2.708
gemma-4-12B4.38 GiB3.148
Qwen2.5-Math-1.5B-Instruct0.60 GiB3.317
Meta-Llama-3.1-8B-Instruct-abliterated2.78 GiB2.978
Phi-3-mini-4k-instruct1.24 GiB2.779
openchat-3.6-8b-202405222.78 GiB2.978
DeepSeek-V2-Lite-Chat6.01 GiB3.288
Hermes-3-Llama-3.1-70B22.79 GiB2.775
Qwen3-Coder-REAP-25B-A3B8.01 GiB2.765
Phi-3-medium-128k-instruct4.44 GiB2.734
Ministral-3-8B-Instruct-2512-BF162.96 GiB2.853
From the filewhat these mean

Other quantizations