Is Llama-3.2-1B-Instruct or Llama-3.2-3B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 1.88 GiB.

Which uses less memory at long context?

Llama-3.2-1B-Instruct — its KV cache at 32K context is 3.5x smaller, because of how its attention layers cache.

Model comparison

Llama-3.2-1B-Instruct vs Llama-3.2-3B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 2,019,377,440. At long context the gap widens: Llama-3.2-1B-Instruct's KV cache at 32K is 3.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Llama-3.2-1B-Instruct	Llama-3.2-3B-Instruct
Parameters	1.2B	3.2B
Architecture	llama	llama
Layers	16	28
Native context	131,072	131,072
Mixture of experts	no	no
Quantizations published	39	39
Smallest quantization	0.39 GiB	0.85 GiB
Q4_K_M	0.75 GiB	1.88 GiB
Licence	—	—

KV cache by context

the term that decides long-context viability

Context	Llama-3.2-1B-Instruct	Llama-3.2-3B-Instruct	Ratio
4,096	0.13 GiB	0.44 GiB	3.50×
8,192	0.25 GiB	0.88 GiB	3.50×
16,384	0.50 GiB	1.75 GiB	3.50×
32,768	1.00 GiB	3.50 GiB	3.50×
65,536	2.00 GiB	7.00 GiB	3.50×
131,072	4.00 GiB	14.00 GiB	3.50×