Is gemma-4-12B-it or Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking smaller?

Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking is smaller at Q4_K_M: 2.32 GiB against 6.63 GiB.

Which uses less memory at long context?

Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking — its KV cache at 32K context is 3.1x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it vs Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking

At Q4_K_M, Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking is the smaller download — 2,489,895,200 bytes against 7,121,861,440. At long context the gap widens: Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking's KV cache at 32K is 3.1× 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	Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking
Parameters	12.0B	4.3B
Architecture	gemma4	gemma3
Layers	48	34
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	59	12
Smallest quantization	3.92 GiB	1.61 GiB
Q4_K_M	6.63 GiB	2.32 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it	Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking	Ratio
4,096	0.72 GiB	0.25 GiB	2.90×
8,192	0.97 GiB	0.33 GiB	2.97×
16,384	1.47 GiB	0.48 GiB	3.04×
32,768	2.47 GiB	0.79 GiB	3.11×
65,536	4.47 GiB	1.42 GiB	3.15×
131,072	8.47 GiB	2.67 GiB	3.17×