Is embeddinggemma-300m-qat-q4_0-unquantized or Qwen3-1.7B smaller?

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

Which uses less memory at long context?

embeddinggemma-300m-qat-q4_0-unquantized — its KV cache at 32K context is 24.2x smaller, because of how its attention layers cache.

Model comparison

embeddinggemma-300m-qat-q4_0-unquantized vs Qwen3-1.7B

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: embeddinggemma-300m-qat-q4_0-unquantized's KV cache at 32K is 24.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	embeddinggemma-300m-qat-q4_0-unquantized	Qwen3-1.7B
Parameters	303M	2.0B
Architecture	gemma-embedding	qwen3
Layers	24	28
Native context	2,048	40,960
Mixture of experts	no	no
Quantizations published	2	48
Smallest quantization	0.21 GiB	0.50 GiB
Q4_K_M	—	1.03 GiB
Licence	gemma	—

KV cache by context

the term that decides long-context viability

Context	embeddinggemma-300m-qat-q4_0-unquantized	Qwen3-1.7B	Ratio
4,096	0.04 GiB	0.44 GiB	12.44×
8,192	0.05 GiB	0.88 GiB	17.23×
16,384	0.08 GiB	1.75 GiB	21.33×
32,768	0.14 GiB	3.50 GiB	24.22×
65,536	0.27 GiB	7.00 GiB	25.97×
131,072	0.52 GiB	14.00 GiB	26.95×