Is SmolLM2-360M-Instruct or Llama-3.2-1B-Instruct smaller?

SmolLM2-360M-Instruct is smaller at Q4_K_M: 0.25 GiB against 0.75 GiB.

Which uses less memory at long context?

Llama-3.2-1B-Instruct — its KV cache at 32K context is 1.3x smaller, because of how its attention layers cache.

Model comparison

SmolLM2-360M-Instruct vs Llama-3.2-1B-Instruct

At Q4_K_M, SmolLM2-360M-Instruct is the smaller download — 270,590,560 bytes against 807,690,688. At long context the gap widens: Llama-3.2-1B-Instruct's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	SmolLM2-360M-Instruct	Llama-3.2-1B-Instruct
Parameters	362M	1.2B
Architecture	llama	llama
Layers	32	16
Native context	8,192	131,072
Mixture of experts	no	no
Quantizations published	27	39
Smallest quantization	0.20 GiB	0.39 GiB
Q4_K_M	0.25 GiB	0.75 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	SmolLM2-360M-Instruct	Llama-3.2-1B-Instruct	Ratio
4,096	0.16 GiB	0.13 GiB	1.25×
8,192	0.31 GiB	0.25 GiB	1.25×
16,384	0.63 GiB	0.50 GiB	1.25×
32,768	1.25 GiB	1.00 GiB	1.25×
65,536	2.50 GiB	2.00 GiB	1.25×
131,072	5.00 GiB	4.00 GiB	1.25×