Is Rocinante-XL-16B-v1 or gemma-4-E4B-it smaller?

gemma-4-E4B-it is smaller at Q4_K_M: 4.64 GiB against 9.37 GiB.

Which uses less memory at long context?

gemma-4-E4B-it — its KV cache at 32K context is 13.3x smaller, because of how its attention layers cache.

Model comparison

Rocinante-XL-16B-v1 vs gemma-4-E4B-it

At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 10,063,921,216. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 13.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Rocinante-XL-16B-v1	gemma-4-E4B-it
Parameters	16.1B	8.0B
Architecture	llama	gemma4
Layers	54	42
Native context	131,072	131,072
Mixture of experts	no	no
Quantizations published	48	36
Smallest quantization	3.54 GiB	3.30 GiB
Q4_K_M	9.37 GiB	4.64 GiB
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Rocinante-XL-16B-v1	gemma-4-E4B-it	Ratio
4,096	0.84 GiB	0.12 GiB	6.86×
8,192	1.69 GiB	0.18 GiB	9.49×
16,384	3.38 GiB	0.29 GiB	11.76×
32,768	6.75 GiB	0.51 GiB	13.34×
65,536	13.50 GiB	0.94 GiB	14.31×
131,072	27.00 GiB	1.82 GiB	14.85×