Is Qwen3.5-4B-DFlash or Llama-3.2-1B-Instruct smaller?

Qwen3.5-4B-DFlash is smaller at Q4_K_M: 0.36 GiB against 0.75 GiB.

Which uses less memory at long context?

Qwen3.5-4B-DFlash — its KV cache at 32K context is 4.7x smaller, because of how its attention layers cache.

Model comparison

Qwen3.5-4B-DFlash vs Llama-3.2-1B-Instruct

At Q4_K_M, Qwen3.5-4B-DFlash is the smaller download — 381,456,064 bytes against 807,690,688. At long context the gap widens: Qwen3.5-4B-DFlash's KV cache at 32K is 4.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3.5-4B-DFlash	Llama-3.2-1B-Instruct
Parameters	634M	1.2B
Architecture	dflash	llama
Layers	6	16
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	8	39
Smallest quantization	0.23 GiB	0.39 GiB
Q4_K_M	0.36 GiB	0.75 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	Qwen3.5-4B-DFlash	Llama-3.2-1B-Instruct	Ratio
4,096	0.09 GiB	0.13 GiB	1.33×
8,192	0.12 GiB	0.25 GiB	2.10×
16,384	0.15 GiB	0.50 GiB	3.32×
32,768	0.21 GiB	1.00 GiB	4.70×
65,536	0.34 GiB	2.00 GiB	5.92×
131,072	0.59 GiB	4.00 GiB	6.80×