Model comparison

North-Mini-Code-1.0 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: North-Mini-Code-1.0's KV cache at 32K is 2.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

North-Mini-Code-1.0gemma-4-12B-it-qat-q4_0-unquantized
Parameters30.5B12.0B
Architecturecohere2moegemma4
Layers4948
Native context500,000262,144
Mixture of expertsyes, 128 expertsno
Quantizations published412
Smallest quantization7.93 GiB6.50 GiB
Q4_K_M17.46 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextNorth-Mini-Code-1.0gemma-4-12B-it-qat-q4_0-unquantizedRatio
4,0960.38 GiB0.72 GiB1.88×
8,1920.52 GiB0.97 GiB1.86×
16,3840.72 GiB1.47 GiB2.03×
32,7681.13 GiB2.47 GiB2.19×
65,5361.94 GiB4.47 GiB2.30×
131,0723.57 GiB8.47 GiB2.37×