Is gemma-4-12B-it or GLM-4.1V-9B-Thinking smaller?

GLM-4.1V-9B-Thinking is smaller at Q4_K_M: 5.74 GiB against 6.63 GiB.

Which uses less memory at long context?

GLM-4.1V-9B-Thinking — its KV cache at 32K context is 2.0x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it vs GLM-4.1V-9B-Thinking

At Q4_K_M, GLM-4.1V-9B-Thinking is the smaller download — 6,166,574,432 bytes against 7,121,861,440. At long context the gap widens: GLM-4.1V-9B-Thinking's KV cache at 32K is 2.0× 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	GLM-4.1V-9B-Thinking
Parameters	12.0B	10.3B
Architecture	gemma4	glm4
Layers	48	40
Native context	262,144	65,536
Mixture of experts	no	no
Quantizations published	59	43
Smallest quantization	3.92 GiB	2.98 GiB
Q4_K_M	6.63 GiB	5.74 GiB
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it	GLM-4.1V-9B-Thinking	Ratio
4,096	0.72 GiB	0.16 GiB	4.60×
8,192	0.97 GiB	0.31 GiB	3.10×
16,384	1.47 GiB	0.63 GiB	2.35×
32,768	2.47 GiB	1.25 GiB	1.98×
65,536	4.47 GiB	2.50 GiB	1.79×
131,072	8.47 GiB	5.00 GiB	1.69×