Quantization publisher

lllyasviel

lllyasviel publishes 22 quantizations across 2 models in our index, averaging 6.249 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 22 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
2
Quantizations
22
Models covered
2
Avg effective bpw
6.249
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantlllyasvielvsTheirsDifference
FLUX.1-devQ2_K3.76 GiBunsloth3.75 GiB+0.2%
FLUX.1-devQ4_K_S6.34 GiBunsloth6.33 GiB+0.1%
FLUX.1-devQ5_K_S7.72 GiBunsloth7.71 GiB+0.1%
FLUX.1-devQ6_K9.18 GiBunsloth9.17 GiB+0.1%
FLUX.1-devQ3_K_S4.87 GiBunsloth4.87 GiB+0.2%
FLUX.1-schnellQ3_K_S4.85 GiBunsloth4.85 GiB+0.1%
FLUX.1-schnellQ5_K_S7.70 GiBunsloth7.69 GiB+0.1%
FLUX.1-schnellQ6_K9.16 GiBunsloth9.15 GiB+0.1%
FLUX.1-schnellQ4_K_S6.32 GiBunsloth6.31 GiB+0.1%
FLUX.1-schnellQ2_K3.73 GiBunsloth3.73 GiB+0.2%
FLUX.1-devQ4_17.01 GiBunsloth7.02 GiB-0.1%
FLUX.1-devQ5_07.70 GiBunsloth7.71 GiB-0.1%
FLUX.1-devF1622.17 GiBunsloth22.17 GiB-0.0%
FLUX.1-devQ8_011.84 GiBunsloth11.84 GiB-0.0%
FLUX.1-devQ4_06.32 GiBunsloth6.33 GiB-0.1%
FLUX.1-devQ5_18.39 GiBunsloth8.40 GiB-0.1%
FLUX.1-schnellQ5_07.68 GiBunsloth7.69 GiB-0.1%
FLUX.1-schnellQ5_18.37 GiBunsloth8.38 GiB-0.1%
FLUX.1-schnellF1622.15 GiBunsloth22.15 GiB-0.0%
FLUX.1-schnellQ8_011.82 GiBunsloth11.82 GiB-0.0%
FLUX.1-schnellQ4_06.31 GiBunsloth6.31 GiB-0.1%
FLUX.1-schnellQ4_16.99 GiBunsloth7.00 GiB-0.1%

A quantization label describes a target, not a recipe. Publishers make different choices about which tensors to keep at higher precision, and some apply an importance matrix while others don't — so two files both honestly labelled the same thing can differ measurably in size and in quality.

Models they publish

ModelQuantizationsSmallest
FLUX.1-dev113.76 GiB
FLUX.1-schnell113.73 GiB