Model comparison

granite-embedding-107m-multilingual vs ced-base

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

granite-embedding-107m-multilingualced-base
Parameters107M86M
Architecturebertced
Layers6
Native context514
Mixture of expertsnono
Quantizations published193
Smallest quantization0.11 GiB0.12 GiB
Q4_K_M0.11 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgranite-embedding-107m-multilingualced-baseRatio
4,0960.04 GiB
8,1920.07 GiB
16,3840.14 GiB
32,7680.28 GiB
65,5360.56 GiB
131,0721.13 GiB