Model comparison

jina-reranker-v1-tiny-en vs mxbai-embed-xsmall-v1

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: jina-reranker-v1-tiny-en's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

jina-reranker-v1-tiny-enmxbai-embed-xsmall-v1
Parameters33M24M
Architecturejina-bert-v2bert
Layers46
Native context8,1924,096
Mixture of expertsnono
Quantizations published124
Smallest quantization0.03 GiB0.03 GiB
Q4_K_M0.03 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextjina-reranker-v1-tiny-enmxbai-embed-xsmall-v1Ratio
4,0960.02 GiB0.04 GiB1.50×
8,1920.05 GiB0.07 GiB1.50×
16,3840.09 GiB0.14 GiB1.50×
32,7680.19 GiB0.28 GiB1.50×
65,5360.38 GiB0.56 GiB1.50×
131,0720.75 GiB1.13 GiB1.50×