Is Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B 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 1.4x smaller, because of how its attention layers cache.

Model comparison

Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B 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 1.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	18.4B	12.0B
Architecture	llama	gemma4
Layers	28	48
Native context	131,072	262,144
Mixture of experts	yes, 8 experts	no
Quantizations published	31	2
Smallest quantization	6.56 GiB	6.50 GiB
Q4_K_M	10.54 GiB	—
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.44 GiB	0.72 GiB	1.64×
8,192	0.88 GiB	0.97 GiB	1.11×
16,384	1.75 GiB	1.47 GiB	1.19×
32,768	3.50 GiB	2.47 GiB	1.42×
65,536	7.00 GiB	4.47 GiB	1.57×
131,072	14.00 GiB	8.47 GiB	1.65×