Is olmOCR-2-7B-1025 or Qwen3-4B smaller?

Qwen3-4B is smaller at Q4_K_M: 2.33 GiB against 4.36 GiB.

Which uses less memory at long context?

olmOCR-2-7B-1025 — its KV cache at 32K context is 2.6x smaller, because of how its attention layers cache.

Model comparison

olmOCR-2-7B-1025 vs Qwen3-4B

At Q4_K_M, Qwen3-4B is the smaller download — 2,497,280,256 bytes against 4,683,072,032. At long context the gap widens: olmOCR-2-7B-1025's KV cache at 32K is 2.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	olmOCR-2-7B-1025	Qwen3-4B
Parameters	8.3B	4.0B
Architecture	qwen2vl	qwen3
Layers	28	36
Native context	128,000	40,960
Mixture of experts	no	no
Quantizations published	50	29
Smallest quantization	1.77 GiB	1.01 GiB
Q4_K_M	4.36 GiB	2.33 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	olmOCR-2-7B-1025	Qwen3-4B	Ratio
4,096	0.22 GiB	0.56 GiB	2.57×
8,192	0.44 GiB	1.13 GiB	2.57×
16,384	0.88 GiB	2.25 GiB	2.57×
32,768	1.75 GiB	4.50 GiB	2.57×
65,536	3.50 GiB	9.00 GiB	2.57×
131,072	7.00 GiB	18.00 GiB	2.57×