Quantization publisher
google publishes 4 quantizations across 4 models in our index, averaging 5.966 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
4
Quantizations
4
Models covered
4
Avg effective bpw
5.966
across their files
Same model, same quant label, different bytes
largest disagreements first
| Model | Quant | vs | Theirs | Difference | |
|---|---|---|---|---|---|
| gemma-3-12b-it | Q4_0 | 7.52 GiB | unsloth | 6.43 GiB | +16.9% |
| gemma-3-4b-it | Q4_0 | 2.94 GiB | bartowski | 2.21 GiB | +33.1% |
| gemma-3-1b-it | Q4_0 | 0.93 GiB | unsloth | 0.67 GiB | +39.0% |
| gemma-4-12B-it-qat-q4_0-unquantized | Q4_0 | 6.50 GiB | lmstudio-community | 6.50 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 |
|---|---|---|
| gemma-4-12B-it-qat-q4_0-unquantized | 1 | 6.50 GiB |
| gemma-3-1b-it | 1 | 0.93 GiB |
| gemma-3-4b-it | 1 | 2.94 GiB |
| gemma-3-12b-it | 1 | 7.52 GiB |