Is 14B or gemma-4-E4B-it smaller?

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

Which uses less memory at long context?

gemma-4-E4B-it — its KV cache at 32K context is 49.4x smaller, because of how its attention layers cache.

Model comparison

14B vs gemma-4-E4B-it

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 49.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	14B	gemma-4-E4B-it
Parameters	14.2B	8.0B
Architecture	llama	gemma4
Layers	40	42
Native context	8,192	131,072
Mixture of experts	no	no
Quantizations published	5	36
Smallest quantization	7.62 GiB	3.30 GiB
Q4_K_M	—	4.64 GiB
Licence	wtfpl	apache-2.0

KV cache by context

the term that decides long-context viability

Context	14B	gemma-4-E4B-it	Ratio
4,096	3.13 GiB	0.12 GiB	25.40×
8,192	6.25 GiB	0.18 GiB	35.16×
16,384	12.50 GiB	0.29 GiB	43.54×
32,768	25.00 GiB	0.51 GiB	49.42×
65,536	50.00 GiB	0.94 GiB	53.00×
131,072	100.00 GiB	1.82 GiB	54.99×