Quantization publisher

AesSedai

AesSedai publishes 8 quantizations across 2 models in our index, averaging 3.478 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 5 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
2
Quantizations
8
Models covered
2
Avg effective bpw
3.478
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantAesSedaivsTheirsDifference
MiMo-V2.5-ProIQ4_XS454.99 GiBbartowski508.09 GiB-10.5%
Kimi-K2.5IQ2_S311.72 GiBbartowski264.31 GiB+17.9%
Kimi-K2.5IQ2_XXS262.75 GiBbartowski228.43 GiB+15.0%
MiMo-V2.5-ProIQ2_S297.46 GiBbartowski290.79 GiB+2.3%
MiMo-V2.5-ProQ4_K_M585.99 GiBbartowski580.09 GiB+1.0%

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
Kimi-K2.53262.75 GiB
MiMo-V2.5-Pro5297.46 GiB