Is codegeex4-all-9b 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 8.1x smaller, because of how its attention layers cache.

Model comparison

codegeex4-all-9b 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 8.1× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	codegeex4-all-9b	gemma-4-12B-it-qat-q4_0-unquantized
Parameters	9.4B	12.0B
Architecture	chatglm	gemma4
Layers	40	48
Native context	—	262,144
Mixture of experts	no	no
Quantizations published	53	2
Smallest quantization	2.89 GiB	6.50 GiB
Q4_K_M	5.82 GiB	—
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	codegeex4-all-9b	gemma-4-12B-it-qat-q4_0-unquantized	Ratio
4,096	2.50 GiB	0.72 GiB	3.48×
8,192	5.00 GiB	0.97 GiB	5.16×
16,384	10.00 GiB	1.47 GiB	6.81×
32,768	20.00 GiB	2.47 GiB	8.10×
65,536	40.00 GiB	4.47 GiB	8.95×
131,072	80.00 GiB	8.47 GiB	9.45×