Is LFM2-VL-3B or Qwen3-VL-4B-Instruct smaller?

LFM2-VL-3B is smaller at Q4_K_M: 1.46 GiB against 2.33 GiB.

Which uses less memory at long context?

LFM2-VL-3B — its KV cache at 32K context is 9.0x smaller, because of how its attention layers cache.

Model comparison

LFM2-VL-3B vs Qwen3-VL-4B-Instruct

At Q4_K_M, LFM2-VL-3B is the smaller download — 1,563,668,256 bytes against 2,497,281,568. At long context the gap widens: LFM2-VL-3B's KV cache at 32K is 9.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	LFM2-VL-3B	Qwen3-VL-4B-Instruct
Parameters	3.0B	4.4B
Architecture	lfm2	qwen3vl
Layers	30	36
Native context	128,000	262,144
Mixture of experts	no	no
Quantizations published	7	26
Smallest quantization	1.38 GiB	1.01 GiB
Q4_K_M	1.46 GiB	2.33 GiB
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	LFM2-VL-3B	Qwen3-VL-4B-Instruct	Ratio
4,096	0.06 GiB	0.56 GiB	8.99×
8,192	0.13 GiB	1.13 GiB	9.00×
16,384	0.25 GiB	2.25 GiB	9.00×
32,768	0.50 GiB	4.50 GiB	9.00×
65,536	1.00 GiB	9.00 GiB	9.00×
131,072	2.00 GiB	18.00 GiB	9.00×