Model comparison

Nethena-20B vs gemma-4-12B-it-qat-q4_0-unquantized

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-12B-it-qat-q4_0-unquantized's KV cache at 32K is 15.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Nethena-20Bgemma-4-12B-it-qat-q4_0-unquantized
Parameters20.0B12.0B
Architecturellamagemma4
Layers6248
Native context4,096262,144
Mixture of expertsnono
Quantizations published122
Smallest quantization7.74 GiB6.50 GiB
Q4_K_M11.22 GiB
Licencecc-by-nc-4.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextNethena-20Bgemma-4-12B-it-qat-q4_0-unquantizedRatio
4,0964.84 GiB0.72 GiB6.74×
8,1929.69 GiB0.97 GiB10.00×
16,38419.38 GiB1.47 GiB13.19×
32,76838.75 GiB2.47 GiB15.70×
65,53677.50 GiB4.47 GiB17.34×
131,072155.00 GiB8.47 GiB18.30×