Is AMALIA-9B-0626-DPO or Qwen3-30B-A3B-Thinking-2507 smaller?

AMALIA-9B-0626-DPO is smaller at Q4_K_M: 5.20 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 1.8x smaller, because of how its attention layers cache.

Model comparison

AMALIA-9B-0626-DPO vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, AMALIA-9B-0626-DPO is the smaller download — 5,582,838,528 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	AMALIA-9B-0626-DPO	Qwen3-30B-A3B-Thinking-2507
Parameters	9.2B	30.5B
Architecture	llama	qwen3moe
Layers	42	48
Native context	32,768	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	19	51
Smallest quantization	3.35 GiB	7.05 GiB
Q4_K_M	5.20 GiB	17.28 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	AMALIA-9B-0626-DPO	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	0.66 GiB	0.38 GiB	1.75×
8,192	1.31 GiB	0.75 GiB	1.75×
16,384	2.63 GiB	1.50 GiB	1.75×
32,768	5.25 GiB	3.00 GiB	1.75×
65,536	10.50 GiB	6.00 GiB	1.75×
131,072	21.00 GiB	12.00 GiB	1.75×