Model comparison

Qwen3-VL-4B-Thinking-Unredacted-MAX vs Qwen3-8B

At Q4_K_M, Qwen3-VL-4B-Thinking-Unredacted-MAX is the smaller download — 2,497,282,656 bytes against 5,027,783,488.

From the file· summed bytes, KV per layer

Side by side

Qwen3-VL-4B-Thinking-Unredacted-MAXQwen3-8B
Parameters4.4B8.2B
Architectureqwen3vlqwen3
Layers3636
Native context262,14440,960
Mixture of expertsnono
Quantizations published3651
Smallest quantization0.98 GiB2.12 GiB
Q4_K_M2.33 GiB4.68 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3-VL-4B-Thinking-Unredacted-MAXQwen3-8BRatio
4,0960.56 GiB0.56 GiB1.00×
8,1921.13 GiB1.13 GiB1.00×
16,3842.25 GiB2.25 GiB1.00×
32,7684.50 GiB4.50 GiB1.00×
65,5369.00 GiB9.00 GiB1.00×
131,07218.00 GiB18.00 GiB1.00×
Qwen3-VL-4B-Thinking-Unredacted-MAX vs Qwen3-8B — size, memory and hardware fit — ossmodeldb