Quantization publisher
mmnga-o
mmnga-o publishes 16 quantizations across 2 models in our index, averaging 5.214 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 3 of the pairs we can compare — the same label does not mean the same file.
From the file· summed file bytes
Repositories
2
Quantizations
16
Models covered
2
Avg effective bpw
5.214
across their files
Same model, same quant label, different bytes
largest disagreements first
| Model | Quant | mmnga-o | vs | Theirs | Difference |
|---|---|---|---|---|---|
| llm-jp-4-32b-a3b-thinking | Q4_K_M | 19.93 GiB | ash2813 | 20.04 GiB | -0.6% |
| llm-jp-4-32b-a3b-thinking | Q4_K_M | 19.93 GiB | hiratagoh | 19.93 GiB | -0.0% |
| llm-jp-4-32b-a3b-thinking | Q5_K_M | 22.73 GiB | hiratagoh | 22.73 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
| Model | Quantizations | Smallest |
|---|---|---|
| llm-jp-4-32b-a3b-thinking | 3 | 15.60 GiB |
| llm-jp-4-8b-instruct | 13 | 3.85 GiB |