Model comparison

Qwen3.8-27B vs GLM-4-32B-0414-Korean-Culture

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 19,680,023,072.

From the file· summed bytes, KV per layer

Side by side

Qwen3.8-27BGLM-4-32B-0414-Korean-Culture
Parameters27.8B32.6B
Architectureqwen35glm4
Layers6461
Native context262,14432,768
Mixture of expertsnono
Quantizations published2234
Smallest quantization8.39 GiB6.77 GiB
Q4_K_M15.66 GiB18.33 GiB
Licenceapache-2.0mit

KV cache by context

the term that decides long-context viability
ContextQwen3.8-27BGLM-4-32B-0414-Korean-CultureRatio
4,0960.25 GiB0.24 GiB1.05×
8,1920.50 GiB0.48 GiB1.05×
16,3841.00 GiB0.95 GiB1.05×
32,7682.00 GiB1.91 GiB1.05×
65,5364.00 GiB3.81 GiB1.05×
131,0728.00 GiB7.63 GiB1.05×