Is PaddleOCR-VL-1.5 or gemma-3-1b-it smaller?

PaddleOCR-VL-1.5 is smaller at Q4_K_M: 0.28 GiB against 0.75 GiB.

Which uses less memory at long context?

gemma-3-1b-it — its KV cache at 32K context is 3.8x smaller, because of how its attention layers cache.

Model comparison

PaddleOCR-VL-1.5 vs gemma-3-1b-it

At Q4_K_M, PaddleOCR-VL-1.5 is the smaller download — 300,218,208 bytes against 806,058,240. At long context the gap widens: gemma-3-1b-it's KV cache at 32K is 3.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	PaddleOCR-VL-1.5	gemma-3-1b-it
Parameters	959M	1000M
Architecture	paddleocr	gemma3
Layers	18	26
Native context	131,072	32,768
Mixture of experts	no	no
Quantizations published	22	28
Smallest quantization	0.25 GiB	0.52 GiB
Q4_K_M	0.28 GiB	0.75 GiB
Licence	—	gemma

KV cache by context

the term that decides long-context viability

Context	PaddleOCR-VL-1.5	gemma-3-1b-it	Ratio
4,096	0.07 GiB	0.04 GiB	1.89×
8,192	0.14 GiB	0.05 GiB	2.67×
16,384	0.28 GiB	0.08 GiB	3.35×
32,768	0.56 GiB	0.15 GiB	3.84×
65,536	1.13 GiB	0.27 GiB	4.14×
131,072	2.25 GiB	0.52 GiB	4.31×