Is DeepSeek-OCR or gemma-4-E4B-it smaller?

DeepSeek-OCR is smaller at Q4_K_M: 1.82 GiB against 4.64 GiB.

Which uses less memory at long context?

gemma-4-E4B-it — its KV cache at 32K context is 1.9x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-OCR vs gemma-4-E4B-it

At Q4_K_M, DeepSeek-OCR is the smaller download — 1,950,326,592 bytes against 4,977,171,584. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 1.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-OCR	gemma-4-E4B-it
Parameters	3.3B	8.0B
Architecture	deepseek_vl_v2	gemma4
Layers	12	42
Native context	8,192	131,072
Mixture of experts	yes, 64 experts	no
Quantizations published	11	36
Smallest quantization	1.54 GiB	3.30 GiB
Q4_K_M	1.82 GiB	4.64 GiB
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	DeepSeek-OCR	gemma-4-E4B-it	Ratio
4,096	0.12 GiB	0.12 GiB	1.05×
8,192	0.23 GiB	0.18 GiB	1.32×
16,384	0.47 GiB	0.29 GiB	1.63×
32,768	0.94 GiB	0.51 GiB	1.85×
65,536	1.88 GiB	0.94 GiB	1.99×
131,072	3.75 GiB	1.82 GiB	2.06×