Quantization publisher

RichardErkhov

RichardErkhov publishes 19 quantizations across 1 models in our index, averaging 5.069 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 16 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
1
Quantizations
19
Models covered
1
Avg effective bpw
5.069
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantRichardErkhovvsTheirsDifference
Qwen2.5-Math-7B-InstructIQ4_XS3.96 GiBbartowski3.93 GiB+0.8%
Qwen2.5-Math-7B-InstructQ4_04.13 GiBbartowski4.14 GiB-0.3%
Qwen2.5-Math-7B-InstructQ5_K_S4.95 GiBbartowski4.95 GiB-0.0%
Qwen2.5-Math-7B-InstructQ6_K5.82 GiBbartowski5.82 GiB-0.0%
Qwen2.5-Math-7B-InstructQ8_07.54 GiBbartowski7.54 GiB-0.0%
Qwen2.5-Math-7B-InstructQ2_K2.81 GiBbartowski2.81 GiB-0.0%
Qwen2.5-Math-7B-InstructQ4_K_S4.15 GiBbartowski4.15 GiB-0.0%
Qwen2.5-Math-7B-InstructQ5_K_M5.07 GiBbartowski5.07 GiB-0.0%
Qwen2.5-Math-7B-InstructQ3_K_L3.81 GiBbartowski3.81 GiB-0.0%
Qwen2.5-Math-7B-InstructQ3_K_M3.55 GiBbartowski3.55 GiB-0.0%
Qwen2.5-Math-7B-InstructQ3_K_S3.25 GiBbartowski3.25 GiB-0.0%
Qwen2.5-Math-7B-InstructQ4_K_M4.36 GiBbartowski4.36 GiB-0.0%
Qwen2.5-Math-7B-InstructQ3_K_L3.81 GiBlmstudio-community3.81 GiB-0.0%
Qwen2.5-Math-7B-InstructQ8_07.54 GiBlmstudio-community7.54 GiB-0.0%
Qwen2.5-Math-7B-InstructQ4_K_M4.36 GiBlmstudio-community4.36 GiB-0.0%
Qwen2.5-Math-7B-InstructQ6_K5.82 GiBlmstudio-community5.82 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
Qwen2.5-Math-7B-Instruct192.81 GiB