Is granite-embedding-107m-multilingual or gemma-3-270m-it smaller?

granite-embedding-107m-multilingual is smaller at Q4_K_M: 0.11 GiB against 0.24 GiB.

Which uses less memory at long context?

gemma-3-270m-it — its KV cache at 32K context is 2.6x smaller, because of how its attention layers cache.

Model comparison

granite-embedding-107m-multilingual vs gemma-3-270m-it

At Q4_K_M, granite-embedding-107m-multilingual is the smaller download — 117,011,136 bytes against 253,115,168. At long context the gap widens: gemma-3-270m-it's KV cache at 32K is 2.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	granite-embedding-107m-multilingual	gemma-3-270m-it
Parameters	107M	268M
Architecture	bert	gemma3
Layers	6	18
Native context	514	32,768
Mixture of experts	no	no
Quantizations published	19	39
Smallest quantization	0.11 GiB	0.17 GiB
Q4_K_M	0.11 GiB	0.24 GiB
Licence	apache-2.0	gemma

KV cache by context

the term that decides long-context viability

Context	granite-embedding-107m-multilingual	gemma-3-270m-it	Ratio
4,096	0.04 GiB	0.03 GiB	1.33×
8,192	0.07 GiB	0.04 GiB	1.85×
16,384	0.14 GiB	0.06 GiB	2.29×
32,768	0.28 GiB	0.11 GiB	2.59×
65,536	0.56 GiB	0.20 GiB	2.78×
131,072	1.13 GiB	0.39 GiB	2.89×