Model comparison

Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored vs Qwen3.8-27B

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 20,414,848,064. At long context the gap widens: Qwen3.8-27B'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

Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-UncensoredQwen3.8-27B
Parameters33.6B27.8B
Architectureqwen3moeqwen35
Layers3664
Native context262,144262,144
Mixture of expertsyes, 12 expertsno
Quantizations published3922
Smallest quantization6.50 GiB8.39 GiB
Q4_K_M19.01 GiB15.66 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-UncensoredQwen3.8-27BRatio
4,0960.56 GiB0.25 GiB2.25×
8,1921.13 GiB0.50 GiB2.25×
16,3842.25 GiB1.00 GiB2.25×
32,7684.50 GiB2.00 GiB2.25×
65,5369.00 GiB4.00 GiB2.25×
131,07218.00 GiB8.00 GiB2.25×