Model comparison

pixtral-12b vs gemma-4-12B-it

At Q4_K_M, gemma-4-12B-it is the smaller download — 7,121,861,440 bytes against 7,477,204,608. At long context the gap widens: gemma-4-12B-it's KV cache at 32K is 2.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

pixtral-12bgemma-4-12B-it
Parameters12.7B12.0B
Architecturellamagemma4
Layers4048
Native context1,024,000262,144
Mixture of expertsnono
Quantizations published3259
Smallest quantization3.85 GiB3.92 GiB
Q4_K_M6.96 GiB6.63 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextpixtral-12bgemma-4-12B-itRatio
4,0960.63 GiB0.72 GiB1.15×
8,1921.25 GiB0.97 GiB1.29×
16,3842.50 GiB1.47 GiB1.70×
32,7685.00 GiB2.47 GiB2.03×
65,53610.00 GiB4.47 GiB2.24×
131,07220.00 GiB8.47 GiB2.36×