Model comparison

llm-jp-4-32b-a3b-thinking vs Qwen3.8-27B

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 21,400,590,016.

From the file· summed bytes, KV per layer

Side by side

llm-jp-4-32b-a3b-thinkingQwen3.8-27B
Parameters32.1B27.8B
Architectureqwen3moeqwen35
Layers3264
Native context65,536262,144
Mixture of expertsyes, 128 expertsno
Quantizations published1122
Smallest quantization15.60 GiB8.39 GiB
Q4_K_M19.93 GiB15.66 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextllm-jp-4-32b-a3b-thinkingQwen3.8-27BRatio
4,0960.25 GiB0.25 GiB1.00×
8,1920.50 GiB0.50 GiB1.00×
16,3841.00 GiB1.00 GiB1.00×
32,7682.00 GiB2.00 GiB1.00×
65,5364.00 GiB4.00 GiB1.00×
131,0728.00 GiB8.00 GiB1.00×