Is Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored or Qwen3-Coder-30B-A3B-Instruct smaller?

Qwen3-Coder-30B-A3B-Instruct is smaller at Q4_K_M: 17.28 GiB against 22.83 GiB.

Which uses less memory at long context?

Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored — its KV cache at 32K context is 1.8x smaller, because of how its attention layers cache.

Model comparison

Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored 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 24,515,131,200. At long context the gap widens: Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored	Qwen3-Coder-30B-A3B-Instruct
Parameters	31.6B	30.5B
Architecture	nemotron_h_moe	qwen3moe
Layers	52	48
Native context	262,144	262,144
Mixture of experts	yes, 128 experts	yes, 128 experts
Quantizations published	34	46
Smallest quantization	16.64 GiB	7.46 GiB
Q4_K_M	22.83 GiB	17.28 GiB
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Nemotron-Cascade-2-30B-A3B-heretic-ara-uncensored	Qwen3-Coder-30B-A3B-Instruct	Ratio
4,096	0.20 GiB	0.38 GiB	1.85×
8,192	0.41 GiB	0.75 GiB	1.85×
16,384	0.81 GiB	1.50 GiB	1.85×
32,768	1.63 GiB	3.00 GiB	1.85×
65,536	3.25 GiB	6.00 GiB	1.85×
131,072	6.50 GiB	12.00 GiB	1.85×