Is SmolVLM2-256M-Video-Instruct or gemma-3-270m-it smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

gemma-3-270m-it — its KV cache at 32K context is 6.5x smaller, because of how its attention layers cache.

Model comparison

SmolVLM2-256M-Video-Instruct vs gemma-3-270m-it

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-3-270m-it's KV cache at 32K is 6.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	SmolVLM2-256M-Video-Instruct	gemma-3-270m-it
Parameters	256M	268M
Architecture	llama	gemma3
Layers	30	18
Native context	8,192	32,768
Mixture of experts	no	no
Quantizations published	26	39
Smallest quantization	0.09 GiB	0.17 GiB
Q4_K_M	—	0.24 GiB
Licence	apache-2.0	gemma

KV cache by context

the term that decides long-context viability

Context	SmolVLM2-256M-Video-Instruct	gemma-3-270m-it	Ratio
4,096	0.09 GiB	0.03 GiB	3.33×
8,192	0.18 GiB	0.04 GiB	4.62×
16,384	0.35 GiB	0.06 GiB	5.71×
32,768	0.70 GiB	0.11 GiB	6.49×
65,536	1.41 GiB	0.20 GiB	6.96×
131,072	2.81 GiB	0.39 GiB	7.22×