Is deepseek-coder-1.3b-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 41.5x smaller, because of how its attention layers cache.

Model comparison

deepseek-coder-1.3b-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 41.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	deepseek-coder-1.3b-instruct	embeddinggemma-300m
Parameters	1.3B	303M
Architecture	llama	gemma-embedding
Layers	24	24
Native context	16,384	2,048
Mixture of experts	no	no
Quantizations published	47	10
Smallest quantization	0.40 GiB	0.26 GiB
Q4_K_M	0.81 GiB	—
Licence	other	—

KV cache by context

the term that decides long-context viability

Context	deepseek-coder-1.3b-instruct	embeddinggemma-300m	Ratio
4,096	0.75 GiB	0.04 GiB	21.33×
8,192	1.50 GiB	0.05 GiB	29.54×
16,384	3.00 GiB	0.08 GiB	36.57×
32,768	6.00 GiB	0.14 GiB	41.51×
65,536	12.00 GiB	0.27 GiB	44.52×
131,072	24.00 GiB	0.52 GiB	46.20×