Model comparison

Laguna-XS-2.1 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: Laguna-XS-2.1's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Laguna-XS-2.1gemma-4-12B-it-qat-q4_0-unquantized
Parameters33.4B12.0B
Architecturelagunagemma4
Layers4048
Native context262,144262,144
Mixture of expertsyes, 256 expertsno
Quantizations published312
Smallest quantization8.76 GiB6.50 GiB
Q4_K_M18.88 GiB
Licenceopenmdw-1.1apache-2.0

KV cache by context

the term that decides long-context viability
ContextLaguna-XS-2.1gemma-4-12B-it-qat-q4_0-unquantizedRatio
4,0960.27 GiB0.72 GiB2.63×
8,1920.43 GiB0.97 GiB2.25×
16,3840.74 GiB1.47 GiB1.98×
32,7681.37 GiB2.47 GiB1.81×
65,5362.62 GiB4.47 GiB1.71×
131,0725.12 GiB8.47 GiB1.65×