Is gemma-4-31B-it-DFlash or Qwen3-4B smaller?

gemma-4-31B-it-DFlash is smaller at Q4_K_M: 0.85 GiB against 2.33 GiB.

Which uses less memory at long context?

gemma-4-31B-it-DFlash — its KV cache at 32K context is 27.4x smaller, because of how its attention layers cache.

Model comparison

gemma-4-31B-it-DFlash vs Qwen3-4B

At Q4_K_M, gemma-4-31B-it-DFlash is the smaller download — 911,587,296 bytes against 2,497,280,256. At long context the gap widens: gemma-4-31B-it-DFlash's KV cache at 32K is 27.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-4-31B-it-DFlash	Qwen3-4B
Parameters	1.5B	4.0B
Architecture	dflash	qwen3
Layers	5	36
Native context	262,144	40,960
Mixture of experts	no	no
Quantizations published	12	29
Smallest quantization	0.54 GiB	1.01 GiB
Q4_K_M	0.85 GiB	2.33 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	gemma-4-31B-it-DFlash	Qwen3-4B	Ratio
4,096	0.05 GiB	0.56 GiB	10.29×
8,192	0.07 GiB	1.13 GiB	16.00×
16,384	0.10 GiB	2.25 GiB	22.15×
32,768	0.16 GiB	4.50 GiB	27.43×
65,536	0.29 GiB	9.00 GiB	31.14×
131,072	0.54 GiB	18.00 GiB	33.39×