Is gemma-4-31B-it-DFlash or HyperCLOVAX-SEED-Text-Instruct-1.5B smaller?

gemma-4-31B-it-DFlash is smaller at Q4_K_M: 0.85 GiB against 1.06 GiB.

Which uses less memory at long context?

We cannot resolve the KV layout for one of these models, so we do not publish a comparison.

Model comparison

At Q4_K_M, gemma-4-31B-it-DFlash is the smaller download — 911,587,296 bytes against 1,133,974,368.

From the file· summed bytes, KV per layer

	gemma-4-31B-it-DFlash	HyperCLOVAX-SEED-Text-Instruct-1.5B
Parameters	1.5B	1.6B
Architecture	dflash	llama
Layers	5	—
Native context	262,144	—
Mixture of experts	no	no
Quantizations published	12	1
Smallest quantization	0.54 GiB	1.06 GiB
Q4_K_M	0.85 GiB	1.06 GiB
Licence	apache-2.0	other

the term that decides long-context viability

Context	gemma-4-31B-it-DFlash	HyperCLOVAX-SEED-Text-Instruct-1.5B	Ratio
4,096	0.05 GiB	—	—
8,192	0.07 GiB	—	—
16,384	0.10 GiB	—	—
32,768	0.16 GiB	—	—
65,536	0.29 GiB	—	—
131,072	0.54 GiB	—	—