Model comparison

gemma-4-12B-it vs gemma-3n-E4B-it

At Q4_K_M, gemma-3n-E4B-it is the smaller download — 4,237,064,064 bytes against 7,121,861,440. At long context the gap widens: gemma-3n-E4B-it's KV cache at 32K is 5.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

gemma-4-12B-itgemma-3n-E4B-it
Parameters12.0B7.8B
Architecturegemma4gemma3n
Layers4835
Native context262,14432,768
Mixture of expertsnono
Quantizations published5938
Smallest quantization3.92 GiB2.57 GiB
Q4_K_M6.63 GiB3.95 GiB
Licenceapache-2.0gemma

KV cache by context

the term that decides long-context viability
Contextgemma-4-12B-itgemma-3n-E4B-itRatio
4,0960.72 GiB0.11 GiB6.57×
8,1920.97 GiB0.16 GiB5.90×
16,3841.47 GiB0.27 GiB5.37×
32,7682.47 GiB0.49 GiB5.02×
65,5364.47 GiB0.93 GiB4.81×
131,0728.47 GiB1.80 GiB4.69×