Is deepseek-coder-1.3b-base or Llama-3.2-1B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 0.81 GiB.

Which uses less memory at long context?

Llama-3.2-1B-Instruct — its KV cache at 32K context is 6.0x smaller, because of how its attention layers cache.

Model comparison

deepseek-coder-1.3b-base vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 873,453,568. At long context the gap widens: Llama-3.2-1B-Instruct's KV cache at 32K is 6.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	deepseek-coder-1.3b-base	Llama-3.2-1B-Instruct
Parameters	1.3B	1.2B
Architecture	llama	llama
Layers	24	16
Native context	16,384	131,072
Mixture of experts	no	no
Quantizations published	49	39
Smallest quantization	0.40 GiB	0.39 GiB
Q4_K_M	0.81 GiB	0.75 GiB
Licence	other	—

KV cache by context

the term that decides long-context viability

Context	deepseek-coder-1.3b-base	Llama-3.2-1B-Instruct	Ratio
4,096	0.75 GiB	0.13 GiB	6.00×
8,192	1.50 GiB	0.25 GiB	6.00×
16,384	3.00 GiB	0.50 GiB	6.00×
32,768	6.00 GiB	1.00 GiB	6.00×
65,536	12.00 GiB	2.00 GiB	6.00×
131,072	24.00 GiB	4.00 GiB	6.00×