Quantization publisher

juiceb0xc0de

juiceb0xc0de publishes 18 quantizations across 2 models in our index, averaging 6.329 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 7 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
2
Quantizations
18
Models covered
2
Avg effective bpw
6.329
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantjuiceb0xc0devsTheirsDifference
llama-3.2-3b-instructQ8_03.19 GiBmfielding923.19 GiB-0.0%
llama-3.2-3b-instructQ4_K_M1.88 GiBZynerji1.88 GiB-0.0%
llama-3.2-3b-instructQ5_K_M2.16 GiBZynerji2.16 GiB-0.0%
llama-3.2-3b-instructQ6_K2.46 GiBZynerji2.46 GiB-0.0%
llama-3.2-3b-instructQ8_03.19 GiBZynerji3.19 GiB-0.0%
llama-3.2-3b-instructIQ3_M1.49 GiBZynerji1.49 GiB-0.0%
llama-3.2-3b-instructIQ4_XS1.70 GiBZynerji1.70 GiB-0.0%

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
Meta-Llama-3.1-8B-Instruct92.43 GiB
llama-3.2-3b-instruct91.02 GiB