Is LFM2-VL-3B or gemma-4-E2B-it smaller?

LFM2-VL-3B is smaller at Q4_K_M: 1.46 GiB against 2.89 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

LFM2-VL-3B vs gemma-4-E2B-it

At Q4_K_M, LFM2-VL-3B is the smaller download — 1,563,668,256 bytes against 3,106,738,272. 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

	LFM2-VL-3B	gemma-4-E2B-it
Parameters	3.0B	5.1B
Architecture	lfm2	gemma4
Layers	30	35
Native context	128,000	131,072
Mixture of experts	no	no
Quantizations published	7	54
Smallest quantization	1.38 GiB	2.13 GiB
Q4_K_M	1.46 GiB	2.89 GiB
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	LFM2-VL-3B	gemma-4-E2B-it	Ratio
4,096	0.06 GiB	0.05 GiB	1.14×
8,192	0.13 GiB	0.08 GiB	1.52×
16,384	0.25 GiB	0.14 GiB	1.83×
32,768	0.50 GiB	0.25 GiB	2.03×
65,536	1.00 GiB	0.46 GiB	2.15×
131,072	2.00 GiB	0.90 GiB	2.22×