Is all-MiniLM-L6-v2 or jina-reranker-v1-tiny-en smaller?

all-MiniLM-L6-v2 is smaller at Q4_K_M: 0.02 GiB against 0.03 GiB.

Which uses less memory at long context?

jina-reranker-v1-tiny-en — its KV cache at 32K context is 1.5x smaller, because of how its attention layers cache.

Model comparison

all-MiniLM-L6-v2 vs jina-reranker-v1-tiny-en

At Q4_K_M, all-MiniLM-L6-v2 is the smaller download — 20,999,104 bytes against 33,257,536. 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

	all-MiniLM-L6-v2	jina-reranker-v1-tiny-en
Parameters	23M	33M
Architecture	bert	jina-bert-v2
Layers	6	4
Native context	512	8,192
Mixture of experts	no	no
Quantizations published	28	12
Smallest quantization	0.02 GiB	0.03 GiB
Q4_K_M	0.02 GiB	0.03 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	all-MiniLM-L6-v2	jina-reranker-v1-tiny-en	Ratio
4,096	0.04 GiB	0.02 GiB	1.50×
8,192	0.07 GiB	0.05 GiB	1.50×
16,384	0.14 GiB	0.09 GiB	1.50×
32,768	0.28 GiB	0.19 GiB	1.50×
65,536	0.56 GiB	0.38 GiB	1.50×
131,072	1.13 GiB	0.75 GiB	1.50×