Is gemma-4-12B-it-qat-q4_0-unquantized or Qwen2.5-VL-32B-Instruct 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 3.2x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs Qwen2.5-VL-32B-Instruct

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 3.2× 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	Qwen2.5-VL-32B-Instruct
Parameters	12.0B	33.5B
Architecture	gemma4	qwen2vl
Layers	48	64
Native context	262,144	128,000
Mixture of experts	no	no
Quantizations published	2	33
Smallest quantization	6.50 GiB	8.41 GiB
Q4_K_M	—	18.49 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	Qwen2.5-VL-32B-Instruct	Ratio
4,096	0.72 GiB	1.00 GiB	1.39×
8,192	0.97 GiB	2.00 GiB	2.06×
16,384	1.47 GiB	4.00 GiB	2.72×
32,768	2.47 GiB	8.00 GiB	3.24×
65,536	4.47 GiB	16.00 GiB	3.58×
131,072	8.47 GiB	32.00 GiB	3.78×