Is PaddleOCR-VL-1.6 or gemma-3-270m-it smaller?

gemma-3-270m-it is smaller at Q4_K_M: 0.24 GiB against 0.28 GiB.

Which uses less memory at long context?

gemma-3-270m-it — its KV cache at 32K context is 5.2x smaller, because of how its attention layers cache.

Model comparison

PaddleOCR-VL-1.6 vs gemma-3-270m-it

At Q4_K_M, gemma-3-270m-it is the smaller download — 253,115,168 bytes against 300,219,456. At long context the gap widens: gemma-3-270m-it's KV cache at 32K is 5.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	PaddleOCR-VL-1.6	gemma-3-270m-it
Parameters	959M	268M
Architecture	paddleocr	gemma3
Layers	18	18
Native context	131,072	32,768
Mixture of experts	no	no
Quantizations published	36	39
Smallest quantization	0.15 GiB	0.17 GiB
Q4_K_M	0.28 GiB	0.24 GiB
Licence	apache-2.0	gemma

KV cache by context

the term that decides long-context viability

Context	PaddleOCR-VL-1.6	gemma-3-270m-it	Ratio
4,096	0.07 GiB	0.03 GiB	2.67×
8,192	0.14 GiB	0.04 GiB	3.69×
16,384	0.28 GiB	0.06 GiB	4.57×
32,768	0.56 GiB	0.11 GiB	5.19×
65,536	1.13 GiB	0.20 GiB	5.57×
131,072	2.25 GiB	0.39 GiB	5.77×