Is gemma-4-12B-it or Muse-Glimmer-30B smaller?

gemma-4-12B-it is smaller at Q4_K_M: 6.63 GiB against 16.12 GiB.

Which uses less memory at long context?

Muse-Glimmer-30B — its KV cache at 32K context is 4.9x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it vs Muse-Glimmer-30B

At Q4_K_M, gemma-4-12B-it is the smaller download — 7,121,861,440 bytes against 17,306,324,000. At long context the gap widens: Muse-Glimmer-30B's KV cache at 32K is 4.9× 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	Muse-Glimmer-30B
Parameters	12.0B	29.8B
Architecture	gemma4	muse-glimmer
Layers	48	52
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	59	36
Smallest quantization	3.92 GiB	8.31 GiB
Q4_K_M	6.63 GiB	16.12 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it	Muse-Glimmer-30B	Ratio
4,096	0.72 GiB	0.15 GiB	4.92×
8,192	0.97 GiB	0.20 GiB	4.92×
16,384	1.47 GiB	0.30 GiB	4.92×
32,768	2.47 GiB	0.50 GiB	4.92×
65,536	4.47 GiB	0.91 GiB	4.92×
131,072	8.47 GiB	1.72 GiB	4.92×