Is Delphi-25B-SimpleRL-Math or Qwen3-30B-A3B-Thinking-2507 smaller?

Delphi-25B-SimpleRL-Math is smaller at Q4_K_M: 14.12 GiB against 17.28 GiB.

Which uses less memory at long context?

Qwen3-30B-A3B-Thinking-2507 — its KV cache at 32K context is 11.2x smaller, because of how its attention layers cache.

Model comparison

Delphi-25B-SimpleRL-Math vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, Delphi-25B-SimpleRL-Math is the smaller download — 15,160,698,624 bytes against 18,556,685,824. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 11.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Delphi-25B-SimpleRL-Math	Qwen3-30B-A3B-Thinking-2507
Parameters	25.0B	30.5B
Architecture	qwen3	qwen3moe
Layers	51	48
Native context	16,384	262,144
Mixture of experts	no	yes, 128 experts
Quantizations published	34	51
Smallest quantization	5.31 GiB	7.05 GiB
Q4_K_M	14.12 GiB	17.28 GiB
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Delphi-25B-SimpleRL-Math	Qwen3-30B-A3B-Thinking-2507	Ratio
4,096	4.18 GiB	0.38 GiB	11.16×
8,192	8.37 GiB	0.75 GiB	11.16×
16,384	16.73 GiB	1.50 GiB	11.16×
32,768	33.47 GiB	3.00 GiB	11.16×
65,536	66.94 GiB	6.00 GiB	11.16×
131,072	133.88 GiB	12.00 GiB	11.16×