Is Ministral-3-8B-Instruct-2512-BF16 or Qwen3-30B-A3B-Thinking-2507 smaller?

Ministral-3-8B-Instruct-2512-BF16 is smaller at Q4_K_M: 4.84 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 1.4x smaller, because of how its attention layers cache.

Model comparison

Ministral-3-8B-Instruct-2512-BF16 vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, Ministral-3-8B-Instruct-2512-BF16 is the smaller download — 5,198,387,168 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 1.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Ministral-3-8B-Instruct-2512-BF16	Qwen3-30B-A3B-Thinking-2507
Parameters	8.9B	30.5B
Architecture	mistral3	qwen3moe
Layers	34	48
Native context	262,144	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	45	51
Smallest quantization	2.89 GiB	7.05 GiB
Q4_K_M	4.84 GiB	17.28 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Ministral-3-8B-Instruct-2512-BF16	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	0.53 GiB	0.38 GiB	1.42×
8,192	1.06 GiB	0.75 GiB	1.42×
16,384	2.13 GiB	1.50 GiB	1.42×
32,768	4.25 GiB	3.00 GiB	1.42×
65,536	8.50 GiB	6.00 GiB	1.42×
131,072	17.00 GiB	12.00 GiB	1.42×