Is gemma-4-12B-it-qat-q4_0-unquantized or gemma-4-E2B-it-heretic-ara smaller?

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

Which uses less memory at long context?

gemma-4-E2B-it-heretic-ara — its KV cache at 32K context is 10.0x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs gemma-4-E2B-it-heretic-ara

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-E2B-it-heretic-ara's KV cache at 32K is 10.0× 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-qat-q4_0-unquantized	gemma-4-E2B-it-heretic-ara
Parameters	12.0B	5.1B
Architecture	gemma4	gemma4
Layers	48	35
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	2	12
Smallest quantization	6.50 GiB	2.78 GiB
Q4_K_M	—	3.19 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	gemma-4-E2B-it-heretic-ara	Ratio
4,096	0.72 GiB	0.05 GiB	13.14×
8,192	0.97 GiB	0.08 GiB	11.81×
16,384	1.47 GiB	0.14 GiB	10.74×
32,768	2.47 GiB	0.25 GiB	10.03×
65,536	4.47 GiB	0.46 GiB	9.61×
131,072	8.47 GiB	0.90 GiB	9.39×