Is granite-embedding-107m-multilingual or SmolLM2-135M-Instruct smaller?

SmolLM2-135M-Instruct is smaller at Q4_K_M: 0.10 GiB against 0.11 GiB.

Which uses less memory at long context?

granite-embedding-107m-multilingual — its KV cache at 32K context is 2.5x smaller, because of how its attention layers cache.

Model comparison

granite-embedding-107m-multilingual vs SmolLM2-135M-Instruct

At Q4_K_M, SmolLM2-135M-Instruct is the smaller download — 105,454,144 bytes against 117,011,136. At long context the gap widens: granite-embedding-107m-multilingual's KV cache at 32K is 2.5× 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	SmolLM2-135M-Instruct
Parameters	107M	135M
Architecture	bert	llama
Layers	6	30
Native context	514	8,192
Mixture of experts	no	no
Quantizations published	19	41
Smallest quantization	0.11 GiB	0.08 GiB
Q4_K_M	0.11 GiB	0.10 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	granite-embedding-107m-multilingual	SmolLM2-135M-Instruct	Ratio
4,096	0.04 GiB	0.09 GiB	2.50×
8,192	0.07 GiB	0.18 GiB	2.50×
16,384	0.14 GiB	0.35 GiB	2.50×
32,768	0.28 GiB	0.70 GiB	2.50×
65,536	0.56 GiB	1.41 GiB	2.50×
131,072	1.13 GiB	2.81 GiB	2.50×