Is gemma-3n-E4B-it or gemma-4-E2B-it smaller?

gemma-4-E2B-it is smaller at Q4_K_M: 2.89 GiB against 3.95 GiB.

Which uses less memory at long context?

gemma-4-E2B-it — its KV cache at 32K context is 2.0x smaller, because of how its attention layers cache.

Model comparison

gemma-3n-E4B-it vs gemma-4-E2B-it

At Q4_K_M, gemma-4-E2B-it is the smaller download — 3,106,738,272 bytes against 4,237,064,064. At long context the gap widens: gemma-4-E2B-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

	gemma-3n-E4B-it	gemma-4-E2B-it
Parameters	7.8B	5.1B
Architecture	gemma3n	gemma4
Layers	35	35
Native context	32,768	131,072
Mixture of experts	no	no
Quantizations published	38	54
Smallest quantization	2.57 GiB	2.13 GiB
Q4_K_M	3.95 GiB	2.89 GiB
Licence	gemma	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-3n-E4B-it	gemma-4-E2B-it	Ratio
4,096	0.11 GiB	0.05 GiB	2.00×
8,192	0.16 GiB	0.08 GiB	2.00×
16,384	0.27 GiB	0.14 GiB	2.00×
32,768	0.49 GiB	0.25 GiB	2.00×
65,536	0.93 GiB	0.46 GiB	2.00×
131,072	1.80 GiB	0.90 GiB	2.00×