Quantization publisher

speakleash

speakleash publishes 14 quantizations across 3 models in our index, averaging 8.380 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 4 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
3
Quantizations
14
Models covered
3
Avg effective bpw
8.380
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantspeakleashvsTheirsDifference
Bielik-11B-v2.2-InstructQ4_K_M6.26 GiBbartowski6.26 GiB-0.0%
Bielik-11B-v2.2-InstructQ5_K_M7.36 GiBbartowski7.36 GiB-0.0%
Bielik-11B-v2.2-InstructQ6_K8.53 GiBbartowski8.53 GiB-0.0%
Bielik-11B-v2.2-InstructQ8_011.05 GiBbartowski11.05 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
Bielik-11B-v3.0-Instruct66.26 GiB
Bielik-11B-v2.3-Instruct46.26 GiB
Bielik-11B-v2.2-Instruct46.26 GiB