Is ZAYA1-8B 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?

ZAYA1-8B — its KV cache at 32K context is 64716.8x smaller, because of how its attention layers cache.

Model comparison

ZAYA1-8B 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: ZAYA1-8B's KV cache at 32K is 64716.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	ZAYA1-8B	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	8.8B	12.0B
Architecture	zaya	gemma4
Layers	40	48
Native context	131,072	262,144
Mixture of experts	yes, 16 experts	no
Quantizations published	6	2
Smallest quantization	4.20 GiB	6.50 GiB
Q4_K_M	5.19 GiB	—
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.72 GiB	18841.60×
8,192	0.97 GiB	25395.20×
16,384	1.47 GiB	38502.40×
32,768	2.47 GiB	64716.80×
65,536	4.47 GiB	117145.60×
131,072	8.47 GiB	222003.20×