Is AnomalyThink-Qwen2.5-VL-7B or Qwen3-8B smaller?

AnomalyThink-Qwen2.5-VL-7B is smaller at Q4_K_M: 4.36 GiB against 4.68 GiB.

Which uses less memory at long context?

AnomalyThink-Qwen2.5-VL-7B — its KV cache at 32K context is 2.6x smaller, because of how its attention layers cache.

Model comparison

AnomalyThink-Qwen2.5-VL-7B vs Qwen3-8B

At Q4_K_M, AnomalyThink-Qwen2.5-VL-7B is the smaller download — 4,683,072,960 bytes against 5,027,783,488. At long context the gap widens: AnomalyThink-Qwen2.5-VL-7B'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

	AnomalyThink-Qwen2.5-VL-7B	Qwen3-8B
Parameters	8.3B	8.2B
Architecture	qwen2vl	qwen3
Layers	28	36
Native context	128,000	40,960
Mixture of experts	no	no
Quantizations published	12	51
Smallest quantization	2.81 GiB	2.12 GiB
Q4_K_M	4.36 GiB	4.68 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	AnomalyThink-Qwen2.5-VL-7B	Qwen3-8B	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×