Is gemma-3-4b-it-abliterated or Qwen2.5-VL-7B-Instruct smaller?

gemma-3-4b-it-abliterated is smaller at Q4_K_M: 2.32 GiB against 4.36 GiB.

Which uses less memory at long context?

gemma-3-4b-it-abliterated — its KV cache at 32K context is 2.2x smaller, because of how its attention layers cache.

Model comparison

gemma-3-4b-it-abliterated vs Qwen2.5-VL-7B-Instruct

At Q4_K_M, gemma-3-4b-it-abliterated is the smaller download — 2,489,894,304 bytes against 4,683,072,032. At long context the gap widens: gemma-3-4b-it-abliterated's KV cache at 32K is 2.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-3-4b-it-abliterated	Qwen2.5-VL-7B-Instruct
Parameters	4.3B	8.3B
Architecture	gemma3	qwen2vl
Layers	34	28
Native context	131,072	128,000
Mixture of experts	no	no
Quantizations published	35	50
Smallest quantization	1.43 GiB	1.93 GiB
Q4_K_M	2.32 GiB	4.36 GiB
Licence	gemma	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-3-4b-it-abliterated	Qwen2.5-VL-7B-Instruct	Ratio
4,096	0.25 GiB	0.22 GiB	1.13×
8,192	0.33 GiB	0.44 GiB	1.34×
16,384	0.48 GiB	0.88 GiB	1.81×
32,768	0.79 GiB	1.75 GiB	2.20×
65,536	1.42 GiB	3.50 GiB	2.46×
131,072	2.67 GiB	7.00 GiB	2.62×