Quantization format

Q5_1

Q5_1 has a nominal rate of bits per weight. We have no measured files carrying this label yet.

From the file· 268 files measured
Nominal bpw
from the block layout
Measured average
6.942
268 files
Range
1.685–31.474
varies by architecture
File sizes
0.03 GiB+
up to 63.57 GiB

Real files

one per model, most downloaded first
ModelSizeEffective bpwvs nominal
FLUX.2-klein-9B6.75 GiB6.390
cohere-transcribe-03-20261.73 GiB7.176
LTX-2.315.21 GiB14.204
whisper-medium0.55 GiB6.143
Wan2.2-I2V-A14B10.26 GiB6.167
Wan2.1-T2V-1.3B1.03 GiB6.206
Jan-v3-4B-base-instruct3.11 GiB6.062
whisper-large-v31.10 GiB6.101
LTX-213.62 GiB6.196
whisper-large-v3-turbo0.58 GiB6.172
granite-4.1-3b2.40 GiB6.052
Qwen-Image-Edit-251114.33 GiB6.027
Krea-2-Turbo9.00 GiB6.032
FLUX.1-dev8.39 GiB6.056
FLUX.2-klein-4B2.93 GiB6.503
Z-Image-Turbo5.15 GiB7.186
LTX2.3-10Eros15.07 GiB6.164
Chroma6.56 GiB6.333
FLUX.1-schnell8.37 GiB6.048
Wan2.2-T2V-A14B10.26 GiB6.167
FLUX.2-dev23.46 GiB6.253
Qwen-Image-251214.33 GiB6.027
Mistral-7B-Instruct-v0.35.07 GiB6.012
Meta-Llama-3-8B-Instruct5.65 GiB6.045
Qwen-Image-Edit-Rapid-AIO14.54 GiB6.114
From the filewhat these mean

Other quantizations