Is granite-4.0-h-small or gemma-4-12B-it-qat-q4_0-unquantized smaller?

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

Which uses less memory at long context?

granite-4.0-h-small — its KV cache at 32K context is 4.9x smaller, because of how its attention layers cache.

Model comparison

granite-4.0-h-small vs gemma-4-12B-it-qat-q4_0-unquantized

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: granite-4.0-h-small'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

	granite-4.0-h-small	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	32.2B	12.0B
Architecture	granitehybrid	gemma4
Layers	40	48
Native context	131,072	262,144
Mixture of experts	yes, 72 experts	no
Quantizations published	41	2
Smallest quantization	7.59 GiB	6.50 GiB
Q4_K_M	18.14 GiB	—
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	granite-4.0-h-small	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.06 GiB	0.72 GiB	11.47×
8,192	0.13 GiB	0.97 GiB	7.74×
16,384	0.25 GiB	1.47 GiB	5.87×
32,768	0.50 GiB	2.47 GiB	4.94×
65,536	1.00 GiB	4.47 GiB	4.47×
131,072	2.00 GiB	8.47 GiB	4.23×