Is Qwen3.6-27B or CodeLlama-13b-Instruct-hf smaller?

CodeLlama-13b-Instruct-hf is smaller at Q4_K_M: 7.33 GiB against 15.41 GiB.

Which uses less memory at long context?

Qwen3.6-27B — its KV cache at 32K context is 12.5x smaller, because of how its attention layers cache.

Model comparison

Qwen3.6-27B vs CodeLlama-13b-Instruct-hf

At Q4_K_M, CodeLlama-13b-Instruct-hf is the smaller download — 7,866,070,016 bytes against 16,547,397,888. At long context the gap widens: Qwen3.6-27B's KV cache at 32K is 12.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3.6-27B	CodeLlama-13b-Instruct-hf
Parameters	27.8B	13.0B
Architecture	qwen35	llama
Layers	64	40
Native context	262,144	16,384
Mixture of experts	no	no
Quantizations published	40	24
Smallest quantization	8.74 GiB	5.06 GiB
Q4_K_M	15.41 GiB	7.33 GiB
Licence	apache-2.0	llama2

KV cache by context

the term that decides long-context viability

Context	Qwen3.6-27B	CodeLlama-13b-Instruct-hf	Ratio
4,096	0.25 GiB	3.13 GiB	12.49×
8,192	0.50 GiB	6.25 GiB	12.50×
16,384	1.00 GiB	12.50 GiB	12.50×
32,768	2.00 GiB	25.00 GiB	12.50×
65,536	4.00 GiB	50.00 GiB	12.50×
131,072	8.00 GiB	100.00 GiB	12.50×