Is Phi-4-reasoning or Qwen3-30B-A3B-Thinking-2507 smaller?

Phi-4-reasoning is smaller at Q4_K_M: 8.43 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 2.1x smaller, because of how its attention layers cache.

Model comparison

Phi-4-reasoning vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, Phi-4-reasoning is the smaller download — 9,053,116,096 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 2.1× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Phi-4-reasoning	Qwen3-30B-A3B-Thinking-2507
Parameters	14.7B	30.5B
Architecture	phi3	qwen3moe
Layers	40	48
Native context	32,768	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	46	51
Smallest quantization	3.33 GiB	7.05 GiB
Q4_K_M	8.43 GiB	17.28 GiB
Licence	mit	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Phi-4-reasoning	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	0.78 GiB	0.38 GiB	2.08×
8,192	1.56 GiB	0.75 GiB	2.08×
16,384	3.13 GiB	1.50 GiB	2.08×
32,768	6.25 GiB	3.00 GiB	2.08×
65,536	12.50 GiB	6.00 GiB	2.08×
131,072	25.00 GiB	12.00 GiB	2.08×