Is deepseek-coder-33b-base 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?

gemma-4-12B-it-qat-q4_0-unquantized — its KV cache at 32K context is 3.1x smaller, because of how its attention layers cache.

Model comparison

deepseek-coder-33b-base 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: gemma-4-12B-it-qat-q4_0-unquantized'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

	deepseek-coder-33b-base	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	33.3B	12.0B
Architecture	llama	gemma4
Layers	62	48
Native context	16,384	262,144
Mixture of experts	no	no
Quantizations published	12	2
Smallest quantization	13.07 GiB	6.50 GiB
Q4_K_M	18.57 GiB	—
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	deepseek-coder-33b-base	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.97 GiB	0.72 GiB	1.35×
8,192	1.94 GiB	0.97 GiB	2.00×
16,384	3.88 GiB	1.47 GiB	2.64×
32,768	7.75 GiB	2.47 GiB	3.14×
65,536	15.50 GiB	4.47 GiB	3.47×
131,072	31.00 GiB	8.47 GiB	3.66×