Quantization publisher

Melvin56

Melvin56 publishes 26 quantizations across 3 models in our index, averaging 6.481 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 9 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
3
Quantizations
26
Models covered
3
Avg effective bpw
6.481
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantMelvin56vsTheirsDifference
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2F1614.19 GiBmradermacher14.19 GiB-0.0%
Phi-4-mini-instruct-abliteratedQ2_K1.57 GiBtensorblock1.57 GiB0.0%
Phi-4-mini-instruct-abliteratedQ3_K_M1.97 GiBtensorblock1.97 GiB0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q5_K_M5.07 GiBmradermacher5.07 GiB-0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q2_K2.81 GiBmradermacher2.81 GiB-0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q8_07.54 GiBmradermacher7.54 GiB-0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q6_K5.82 GiBmradermacher5.82 GiB-0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q3_K_M3.55 GiBmradermacher3.55 GiB-0.0%
DeepSeek-R1-Distill-Qwen-7B-abliterated-v2Q4_K_M4.36 GiBmradermacher4.36 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
Qwen3-4B-abliterated81.55 GiB
DeepSeek-R1-Distill-Qwen-7B-abliterated-v292.64 GiB
Phi-4-mini-instruct-abliterated91.48 GiB