Is Llama-3_3-Nemotron-Super-49B-v1 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 32.4x smaller, because of how its attention layers cache.

Model comparison

Llama-3_3-Nemotron-Super-49B-v1 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 32.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Llama-3_3-Nemotron-Super-49B-v1	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	49.9B	12.0B
Architecture	deci	gemma4
Layers	80	48
Native context	131,072	262,144
Mixture of experts	no	no
Quantizations published	66	2
Smallest quantization	10.27 GiB	6.50 GiB
Q4_K_M	28.14 GiB	—
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Llama-3_3-Nemotron-Super-49B-v1	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	10.00 GiB	0.72 GiB	13.91×
8,192	20.00 GiB	0.97 GiB	20.65×
16,384	40.00 GiB	1.47 GiB	27.23×
32,768	80.00 GiB	2.47 GiB	32.41×
65,536	160.00 GiB	4.47 GiB	35.80×
131,072	320.00 GiB	8.47 GiB	37.79×