Is Qwen3-Coder-30B-A3B-Instruct or Rocinante-XL-16B-v1 smaller?

Rocinante-XL-16B-v1 is smaller at Q4_K_M: 9.37 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-Coder-30B-A3B-Instruct — its KV cache at 32K context is 2.3x smaller, because of how its attention layers cache.

Model comparison

Qwen3-Coder-30B-A3B-Instruct vs Rocinante-XL-16B-v1

At Q4_K_M, Rocinante-XL-16B-v1 is the smaller download — 10,063,921,216 bytes against 18,556,689,568. At long context the gap widens: Qwen3-Coder-30B-A3B-Instruct's KV cache at 32K is 2.3× 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	Rocinante-XL-16B-v1
Parameters	30.5B	16.1B
Architecture	qwen3moe	llama
Layers	48	54
Native context	262,144	131,072
Mixture of experts	yes, 128 experts	no
Quantizations published	46	48
Smallest quantization	7.46 GiB	3.54 GiB
Q4_K_M	17.28 GiB	9.37 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	Qwen3-Coder-30B-A3B-Instruct	Rocinante-XL-16B-v1	Ratio
4,096	0.38 GiB	0.84 GiB	2.25×
8,192	0.75 GiB	1.69 GiB	2.25×
16,384	1.50 GiB	3.38 GiB	2.25×
32,768	3.00 GiB	6.75 GiB	2.25×
65,536	6.00 GiB	13.50 GiB	2.25×
131,072	12.00 GiB	27.00 GiB	2.25×