Is Skyfall-31B-v4.2 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?

gemma-4-12B-it-qat-q4_0-unquantized — its KV cache at 32K context is 2.7x smaller, because of how its attention layers cache.

Model comparison

Skyfall-31B-v4.2 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: gemma-4-12B-it-qat-q4_0-unquantized's KV cache at 32K is 2.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Skyfall-31B-v4.2	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	31.4B	12.0B
Architecture	llama	gemma4
Layers	54	48
Native context	131,072	262,144
Mixture of experts	no	no
Quantizations published	49	2
Smallest quantization	6.39 GiB	6.50 GiB
Q4_K_M	17.97 GiB	—
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Skyfall-31B-v4.2	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.84 GiB	0.72 GiB	1.17×
8,192	1.69 GiB	0.97 GiB	1.74×
16,384	3.38 GiB	1.47 GiB	2.30×
32,768	6.75 GiB	2.47 GiB	2.73×
65,536	13.50 GiB	4.47 GiB	3.02×
131,072	27.00 GiB	8.47 GiB	3.19×