Is gemma-4-12B-it-qat-q4_0-unquantized or GLM-4-32B-0414-Korean-Culture smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

GLM-4-32B-0414-Korean-Culture — its KV cache at 32K context is 1.3x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs GLM-4-32B-0414-Korean-Culture

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: GLM-4-32B-0414-Korean-Culture's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-4-12B-it-qat-q4_0-unquantized	GLM-4-32B-0414-Korean-Culture
Parameters	12.0B	32.6B
Architecture	gemma4	glm4
Layers	48	61
Native context	262,144	32,768
Mixture of experts	no	no
Quantizations published	2	34
Smallest quantization	6.50 GiB	6.77 GiB
Q4_K_M	—	18.33 GiB
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	GLM-4-32B-0414-Korean-Culture	Ratio
4,096	0.72 GiB	0.24 GiB	3.02×
8,192	0.97 GiB	0.48 GiB	2.03×
16,384	1.47 GiB	0.95 GiB	1.54×
32,768	2.47 GiB	1.91 GiB	1.30×
65,536	4.47 GiB	3.81 GiB	1.17×
131,072	8.47 GiB	7.63 GiB	1.11×