Is gemma-4-12B-it-qat-q4_0-unquantized or Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop — its KV cache at 32K context is 3.3x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop's KV cache at 32K is 3.3× 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	Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop
Parameters	12.0B	12.1B
Architecture	gemma4	qwen35
Layers	48	24
Native context	262,144	262,144
Mixture of experts	no	no
Quantizations published	2	33
Smallest quantization	6.50 GiB	4.60 GiB
Q4_K_M	—	6.82 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	Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop	Ratio
4,096	0.72 GiB	0.09 GiB	7.66×
8,192	0.97 GiB	0.19 GiB	5.16×
16,384	1.47 GiB	0.38 GiB	3.92×
32,768	2.47 GiB	0.75 GiB	3.29×
65,536	4.47 GiB	1.50 GiB	2.98×
131,072	8.47 GiB	3.00 GiB	2.82×