Is gemma-4-12B-it-qat-q4_0-unquantized or deepseek-llm-67b-chat 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 4.8x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs deepseek-llm-67b-chat

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 4.8× 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	deepseek-llm-67b-chat
Parameters	12.0B	67.4B
Architecture	gemma4	llama
Layers	48	95
Native context	262,144	4,096
Mixture of experts	no	no
Quantizations published	2	42
Smallest quantization	6.50 GiB	13.65 GiB
Q4_K_M	—	37.66 GiB
Licence	apache-2.0	other

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	deepseek-llm-67b-chat	Ratio
4,096	0.72 GiB	1.48 GiB	2.07×
8,192	0.97 GiB	2.97 GiB	3.06×
16,384	1.47 GiB	5.94 GiB	4.04×
32,768	2.47 GiB	11.88 GiB	4.81×
65,536	4.47 GiB	23.75 GiB	5.31×
131,072	8.47 GiB	47.50 GiB	5.61×