Is gte-large or embeddinggemma-300m-qat-q8_0-unquantized smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

embeddinggemma-300m-qat-q8_0-unquantized — its KV cache at 32K context is 20.8x smaller, because of how its attention layers cache.

Model comparison

gte-large vs embeddinggemma-300m-qat-q8_0-unquantized

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: embeddinggemma-300m-qat-q8_0-unquantized's KV cache at 32K is 20.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gte-large	embeddinggemma-300m-qat-q8_0-unquantized
Parameters	335M	303M
Architecture	bert	gemma-embedding
Layers	24	24
Native context	512	2,048
Mixture of experts	no	no
Quantizations published	12	1
Smallest quantization	0.13 GiB	0.31 GiB
Q4_K_M	0.20 GiB	—
Licence	mit	gemma

KV cache by context

the term that decides long-context viability

Context	gte-large	embeddinggemma-300m-qat-q8_0-unquantized	Ratio
4,096	0.38 GiB	0.04 GiB	10.67×
8,192	0.75 GiB	0.05 GiB	14.77×
16,384	1.50 GiB	0.08 GiB	18.29×
32,768	3.00 GiB	0.14 GiB	20.76×
65,536	6.00 GiB	0.27 GiB	22.26×
131,072	12.00 GiB	0.52 GiB	23.10×