Is gemma-2b or Llama-3.2-1B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 1.39 GiB.

Which uses less memory at long context?

gemma-2b — its KV cache at 32K context is 1.8x smaller, because of how its attention layers cache.

Model comparison

gemma-2b vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 1,495,245,728. At long context the gap widens: gemma-2b's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-2b	Llama-3.2-1B-Instruct
Parameters	2.5B	1.2B
Architecture	gemma	llama
Layers	18	16
Native context	8,192	131,072
Mixture of experts	no	no
Quantizations published	38	39
Smallest quantization	0.84 GiB	0.39 GiB
Q4_K_M	1.39 GiB	0.75 GiB
Licence	cc-by-nc-4.0	—

KV cache by context

the term that decides long-context viability

Context	gemma-2b	Llama-3.2-1B-Instruct	Ratio
4,096	0.07 GiB	0.13 GiB	1.78×
8,192	0.14 GiB	0.25 GiB	1.78×
16,384	0.28 GiB	0.50 GiB	1.78×
32,768	0.56 GiB	1.00 GiB	1.78×
65,536	1.13 GiB	2.00 GiB	1.78×
131,072	2.25 GiB	4.00 GiB	1.78×