Model comparison

GLM-4-32B-0414 vs Qwen3.6-27B

At Q4_K_M, Qwen3.6-27B is the smaller download — 16,547,397,888 bytes against 19,680,022,112.

From the file· summed bytes, KV per layer

Side by side

GLM-4-32B-0414Qwen3.6-27B
Parameters32.6B27.8B
Architectureglm4qwen35
Layers6164
Native context32,768262,144
Mixture of expertsnono
Quantizations published4940
Smallest quantization7.17 GiB8.74 GiB
Q4_K_M18.33 GiB15.41 GiB
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
ContextGLM-4-32B-0414Qwen3.6-27BRatio
4,0960.24 GiB0.25 GiB1.05×
8,1920.48 GiB0.50 GiB1.05×
16,3840.95 GiB1.00 GiB1.05×
32,7681.91 GiB2.00 GiB1.05×
65,5363.81 GiB4.00 GiB1.05×
131,0727.63 GiB8.00 GiB1.05×