Is Qwen3.8-27B or Llama-3.2-11B-Vision-Instruct smaller?

Llama-3.2-11B-Vision-Instruct is smaller at Q4_K_M: 5.55 GiB against 15.66 GiB.

Which uses less memory at long context?

Qwen3.8-27B — its KV cache at 32K context is 2.5x smaller, because of how its attention layers cache.

Model comparison

Qwen3.8-27B vs Llama-3.2-11B-Vision-Instruct

At Q4_K_M, Llama-3.2-11B-Vision-Instruct is the smaller download — 5,963,057,216 bytes against 16,810,714,336. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 2.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3.8-27B	Llama-3.2-11B-Vision-Instruct
Parameters	27.8B	10.7B
Architecture	qwen35	mllama
Layers	64	40
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	22	9
Smallest quantization	8.39 GiB	5.55 GiB
Q4_K_M	15.66 GiB	5.55 GiB
Licence	apache-2.0	llama3.2

KV cache by context

the term that decides long-context viability

Context	Qwen3.8-27B	Llama-3.2-11B-Vision-Instruct	Ratio
4,096	0.25 GiB	0.63 GiB	2.50×
8,192	0.50 GiB	1.25 GiB	2.50×
16,384	1.00 GiB	2.50 GiB	2.50×
32,768	2.00 GiB	5.00 GiB	2.50×
65,536	4.00 GiB	10.00 GiB	2.50×
131,072	8.00 GiB	20.00 GiB	2.50×