Is LFM2.5-ColBERT-350M or jina-embeddings-v5-text-nano smaller?

jina-embeddings-v5-text-nano is smaller at Q4_K_M: 0.15 GiB against 0.21 GiB.

Which uses less memory at long context?

LFM2.5-ColBERT-350M — its KV cache at 32K context is 3.0x smaller, because of how its attention layers cache.

Model comparison

LFM2.5-ColBERT-350M vs jina-embeddings-v5-text-nano

At Q4_K_M, jina-embeddings-v5-text-nano is the smaller download — 156,966,976 bytes against 228,407,552. At long context the gap widens: LFM2.5-ColBERT-350M's KV cache at 32K is 3.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	LFM2.5-ColBERT-350M	jina-embeddings-v5-text-nano
Parameters	353M	212M
Architecture	lfm2	eurobert
Layers	16	12
Native context	128,000	8,192
Mixture of experts	no	no
Quantizations published	24	42
Smallest quantization	0.15 GiB	0.09 GiB
Q4_K_M	0.21 GiB	0.15 GiB
Licence	other	cc-by-nc-4.0

KV cache by context

the term that decides long-context viability

Context	LFM2.5-ColBERT-350M	jina-embeddings-v5-text-nano	Ratio
4,096	0.05 GiB	0.14 GiB	3.00×
8,192	0.09 GiB	0.28 GiB	3.00×
16,384	0.19 GiB	0.56 GiB	3.00×
32,768	0.38 GiB	1.13 GiB	3.00×
65,536	0.75 GiB	2.25 GiB	3.00×
131,072	1.50 GiB	4.50 GiB	3.00×