Is Falcon3-1B-Instruct or embeddinggemma-300m smaller?

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

Which uses less memory at long context?

embeddinggemma-300m — its KV cache at 32K context is 15.6x smaller, because of how its attention layers cache.

Model comparison

Falcon3-1B-Instruct vs embeddinggemma-300m

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: embeddinggemma-300m's KV cache at 32K is 15.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Falcon3-1B-Instruct	embeddinggemma-300m
Parameters	1.7B	303M
Architecture	llama	gemma-embedding
Layers	18	24
Native context	8,192	2,048
Mixture of experts	no	no
Quantizations published	30	10
Smallest quantization	0.21 GiB	0.26 GiB
Q4_K_M	0.98 GiB	—
Licence	other	—

KV cache by context

the term that decides long-context viability

Context	Falcon3-1B-Instruct	embeddinggemma-300m	Ratio
4,096	0.28 GiB	0.04 GiB	8.00×
8,192	0.56 GiB	0.05 GiB	11.08×
16,384	1.13 GiB	0.08 GiB	13.71×
32,768	2.25 GiB	0.14 GiB	15.57×
65,536	4.50 GiB	0.27 GiB	16.70×
131,072	9.00 GiB	0.52 GiB	17.32×