Is Phind-CodeLlama-34B-v2 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 2.4x smaller, because of how its attention layers cache.

Model comparison

Phind-CodeLlama-34B-v2 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 2.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Phind-CodeLlama-34B-v2	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	33.7B	12.0B
Architecture	llama	gemma4
Layers	48	48
Native context	16,384	262,144
Mixture of experts	no	no
Quantizations published	33	2
Smallest quantization	6.75 GiB	6.50 GiB
Q4_K_M	18.83 GiB	—
Licence	llama2	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Phind-CodeLlama-34B-v2	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	0.75 GiB	0.72 GiB	1.04×
8,192	1.50 GiB	0.97 GiB	1.55×
16,384	3.00 GiB	1.47 GiB	2.04×
32,768	6.00 GiB	2.47 GiB	2.43×
65,536	12.00 GiB	4.47 GiB	2.69×
131,072	24.00 GiB	8.47 GiB	2.83×