Is Qwen3-Coder-30B-A3B-Instruct or DeepSeek-R1-Distill-Qwen-32B-heretic smaller?

Qwen3-Coder-30B-A3B-Instruct is smaller at Q4_K_M: 17.28 GiB against 18.49 GiB.

Which uses less memory at long context?

Qwen3-Coder-30B-A3B-Instruct — its KV cache at 32K context is 2.7x smaller, because of how its attention layers cache.

Model comparison

Qwen3-Coder-30B-A3B-Instruct vs DeepSeek-R1-Distill-Qwen-32B-heretic

At Q4_K_M, Qwen3-Coder-30B-A3B-Instruct is the smaller download — 18,556,689,568 bytes against 19,851,336,352. At long context the gap widens: Qwen3-Coder-30B-A3B-Instruct's KV cache at 32K is 2.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3-Coder-30B-A3B-Instruct	DeepSeek-R1-Distill-Qwen-32B-heretic
Parameters	30.5B	32.8B
Architecture	qwen3moe	qwen2
Layers	48	64
Native context	262,144	131,072
Mixture of experts	yes, 128 experts	no
Quantizations published	46	57
Smallest quantization	7.46 GiB	6.77 GiB
Q4_K_M	17.28 GiB	18.49 GiB
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	Qwen3-Coder-30B-A3B-Instruct	DeepSeek-R1-Distill-Qwen-32B-heretic	Ratio
4,096	0.38 GiB	1.00 GiB	2.67×
8,192	0.75 GiB	2.00 GiB	2.67×
16,384	1.50 GiB	4.00 GiB	2.67×
32,768	3.00 GiB	8.00 GiB	2.67×
65,536	6.00 GiB	16.00 GiB	2.67×
131,072	12.00 GiB	32.00 GiB	2.67×