Is SmolLM2-CoT-360M or Qwen3-1.7B smaller?

SmolLM2-CoT-360M is smaller at Q4_K_M: 0.25 GiB against 1.03 GiB.

Which uses less memory at long context?

SmolLM2-CoT-360M — its KV cache at 32K context is 2.8x smaller, because of how its attention layers cache.

Model comparison

SmolLM2-CoT-360M vs Qwen3-1.7B

At Q4_K_M, SmolLM2-CoT-360M is the smaller download — 270,591,136 bytes against 1,107,409,472. At long context the gap widens: SmolLM2-CoT-360M's KV cache at 32K is 2.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	SmolLM2-CoT-360M	Qwen3-1.7B
Parameters	362M	2.0B
Architecture	llama	qwen3
Layers	32	28
Native context	8,192	40,960
Mixture of experts	no	no
Quantizations published	1	48
Smallest quantization	0.25 GiB	0.50 GiB
Q4_K_M	0.25 GiB	1.03 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	SmolLM2-CoT-360M	Qwen3-1.7B	Ratio
4,096	0.16 GiB	0.44 GiB	2.80×
8,192	0.31 GiB	0.88 GiB	2.80×
16,384	0.63 GiB	1.75 GiB	2.80×
32,768	1.25 GiB	3.50 GiB	2.80×
65,536	2.50 GiB	7.00 GiB	2.80×
131,072	5.00 GiB	14.00 GiB	2.80×