Is DeepSeek-R1-Distill-Llama-8B-Abliterated or Qwen3-30B-A3B-Thinking-2507 smaller?

DeepSeek-R1-Distill-Llama-8B-Abliterated is smaller at Q4_K_M: 9.17 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 1.3x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-R1-Distill-Llama-8B-Abliterated vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, DeepSeek-R1-Distill-Llama-8B-Abliterated is the smaller download — 9,841,474,624 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 1.3× 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-Abliterated	Qwen3-30B-A3B-Thinking-2507
Parameters	8.0B	30.5B
Architecture	llama	qwen3moe
Layers	32	48
Native context	131,072	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	60	51
Smallest quantization	3.76 GiB	7.05 GiB
Q4_K_M	9.17 GiB	17.28 GiB
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	DeepSeek-R1-Distill-Llama-8B-Abliterated	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	0.50 GiB	0.38 GiB	1.33×
8,192	1.00 GiB	0.75 GiB	1.33×
16,384	2.00 GiB	1.50 GiB	1.33×
32,768	4.00 GiB	3.00 GiB	1.33×
65,536	8.00 GiB	6.00 GiB	1.33×
131,072	16.00 GiB	12.00 GiB	1.33×