Is granite-3.1-1b-a400m-instruct or Llama-3.2-1B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 0.77 GiB.

Which uses less memory at long context?

Llama-3.2-1B-Instruct — its KV cache at 32K context is 1.5x smaller, because of how its attention layers cache.

Model comparison

granite-3.1-1b-a400m-instruct vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 821,847,360. At long context the gap widens: Llama-3.2-1B-Instruct's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	granite-3.1-1b-a400m-instruct	Llama-3.2-1B-Instruct
Parameters	1.3B	1.2B
Architecture	granitemoe	llama
Layers	24	16
Native context	131,072	131,072
Mixture of experts	yes, 32 experts	no
Quantizations published	20	39
Smallest quantization	0.48 GiB	0.39 GiB
Q4_K_M	0.77 GiB	0.75 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	granite-3.1-1b-a400m-instruct	Llama-3.2-1B-Instruct	Ratio
4,096	0.19 GiB	0.13 GiB	1.50×
8,192	0.38 GiB	0.25 GiB	1.50×
16,384	0.75 GiB	0.50 GiB	1.50×
32,768	1.50 GiB	1.00 GiB	1.50×
65,536	3.00 GiB	2.00 GiB	1.50×
131,072	6.00 GiB	4.00 GiB	1.50×