Is Devstral-Small-2507 or Qwen3-30B-A3B-Thinking-2507 smaller?

Devstral-Small-2507 is smaller at Q4_K_M: 13.35 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 1.7x smaller, because of how its attention layers cache.

Model comparison

Devstral-Small-2507 vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, Devstral-Small-2507 is the smaller download — 14,333,915,904 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 1.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Devstral-Small-2507	Qwen3-30B-A3B-Thinking-2507
Parameters	23.6B	30.5B
Architecture	llama	qwen3moe
Layers	40	48
Native context	131,072	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	27	51
Smallest quantization	5.18 GiB	7.05 GiB
Q4_K_M	13.35 GiB	17.28 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Devstral-Small-2507	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	0.63 GiB	0.38 GiB	1.67×
8,192	1.25 GiB	0.75 GiB	1.67×
16,384	2.50 GiB	1.50 GiB	1.67×
32,768	5.00 GiB	3.00 GiB	1.67×
65,536	10.00 GiB	6.00 GiB	1.67×
131,072	20.00 GiB	12.00 GiB	1.67×