Is deepseek-r1-distill-llama-8b-unsloth-bnb-4bit or Qwen3.8-27B smaller?

deepseek-r1-distill-llama-8b-unsloth-bnb-4bit is smaller at Q4_K_M: 14.97 GiB against 15.66 GiB.

Which uses less memory at long context?

Qwen3.8-27B — its KV cache at 32K context is 2.0x smaller, because of how its attention layers cache.

Model comparison

deepseek-r1-distill-llama-8b-unsloth-bnb-4bit vs Qwen3.8-27B

At Q4_K_M, deepseek-r1-distill-llama-8b-unsloth-bnb-4bit is the smaller download — 16,068,894,016 bytes against 16,810,714,336. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 2.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	deepseek-r1-distill-llama-8b-unsloth-bnb-4bit	Qwen3.8-27B
Parameters	8.2B	27.8B
Architecture	llama	qwen35
Layers	32	64
Native context	131,072	262,144
Mixture of experts	no	no
Quantizations published	2	22
Smallest quantization	7.95 GiB	8.39 GiB
Q4_K_M	14.97 GiB	15.66 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	deepseek-r1-distill-llama-8b-unsloth-bnb-4bit	Qwen3.8-27B	Ratio
4,096	0.50 GiB	0.25 GiB	2.00×
8,192	1.00 GiB	0.50 GiB	2.00×
16,384	2.00 GiB	1.00 GiB	2.00×
32,768	4.00 GiB	2.00 GiB	2.00×
65,536	8.00 GiB	4.00 GiB	2.00×
131,072	16.00 GiB	8.00 GiB	2.00×