Is Trinity-Mini 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?

Trinity-Mini — its KV cache at 32K context is 4.0x smaller, because of how its attention layers cache.

Model comparison

Trinity-Mini 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: Trinity-Mini's KV cache at 32K is 4.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Trinity-Mini	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	26.1B	12.0B
Architecture	afmoe	gemma4
Layers	32	48
Native context	131,072	262,144
Mixture of experts	yes, 128 experts	no
Quantizations published	57	2
Smallest quantization	6.11 GiB	6.50 GiB
Q4_K_M	14.74 GiB	—
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Trinity-Mini	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.18 GiB	0.72 GiB	4.00×
8,192	0.24 GiB	0.97 GiB	4.00×
16,384	0.37 GiB	1.47 GiB	4.00×
32,768	0.62 GiB	2.47 GiB	4.00×
65,536	1.12 GiB	4.47 GiB	4.00×
131,072	2.12 GiB	8.47 GiB	4.00×