Is gemma-4-12B-it or gemma-3n-E2B-it smaller?

gemma-3n-E2B-it is smaller at Q4_K_M: 2.60 GiB against 6.63 GiB.

Which uses less memory at long context?

gemma-3n-E2B-it — its KV cache at 32K context is 5.9x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it vs gemma-3n-E2B-it

At Q4_K_M, gemma-3n-E2B-it is the smaller download — 2,787,806,304 bytes against 7,121,861,440. At long context the gap widens: gemma-3n-E2B-it's KV cache at 32K is 5.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-4-12B-it	gemma-3n-E2B-it
Parameters	12.0B	5.4B
Architecture	gemma4	gemma3n
Layers	48	30
Native context	262,144	32,768
Mixture of experts	no	no
Quantizations published	59	40
Smallest quantization	3.92 GiB	1.76 GiB
Q4_K_M	6.63 GiB	2.60 GiB
Licence	apache-2.0	gemma

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it	gemma-3n-E2B-it	Ratio
4,096	0.72 GiB	0.09 GiB	7.67×
8,192	0.97 GiB	0.14 GiB	6.89×
16,384	1.47 GiB	0.23 GiB	6.27×
32,768	2.47 GiB	0.42 GiB	5.85×
65,536	4.47 GiB	0.80 GiB	5.61×
131,072	8.47 GiB	1.55 GiB	5.47×