Is Rombos-LLM-V2.5-Qwen-32b or Qwen3-Coder-30B-A3B-Instruct 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

Rombos-LLM-V2.5-Qwen-32b vs Qwen3-Coder-30B-A3B-Instruct

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

	Rombos-LLM-V2.5-Qwen-32b	Qwen3-Coder-30B-A3B-Instruct
Parameters	32.8B	30.5B
Architecture	qwen2	qwen3moe
Layers	64	48
Native context	32,768	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	23	46
Smallest quantization	9.27 GiB	7.46 GiB
Q4_K_M	18.49 GiB	17.28 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Rombos-LLM-V2.5-Qwen-32b	Qwen3-Coder-30B-A3B-Instruct	Ratio
4,096	1.00 GiB	0.38 GiB	2.67×
8,192	2.00 GiB	0.75 GiB	2.67×
16,384	4.00 GiB	1.50 GiB	2.67×
32,768	8.00 GiB	3.00 GiB	2.67×
65,536	16.00 GiB	6.00 GiB	2.67×
131,072	32.00 GiB	12.00 GiB	2.67×