Is gemma-4-12B-it-qat-q4_0-unquantized or Skyfall-31B-v4.2-heretic 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

gemma-4-12B-it-qat-q4_0-unquantized vs Skyfall-31B-v4.2-heretic

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

	gemma-4-12B-it-qat-q4_0-unquantized	Skyfall-31B-v4.2-heretic
Parameters	12.0B	31.4B
Architecture	gemma4	llama
Layers	48	54
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	2	34
Smallest quantization	6.50 GiB	6.39 GiB
Q4_K_M	—	17.68 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	Skyfall-31B-v4.2-heretic	Ratio
4,096	0.72 GiB	0.84 GiB	1.17×
8,192	0.97 GiB	1.69 GiB	1.74×
16,384	1.47 GiB	3.38 GiB	2.30×
32,768	2.47 GiB	6.75 GiB	2.73×
65,536	4.47 GiB	13.50 GiB	3.02×
131,072	8.47 GiB	27.00 GiB	3.19×