Model comparison

Qwen3.8-27B vs Llama3.2-30B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-Uncensored

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 18,366,980,192. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 2.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.8-27BLlama3.2-30B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-Uncensored
Parameters27.8B30.0B
Architectureqwen35llama
Layers6447
Native context262,144131,072
Mixture of expertsnoyes, 8 experts
Quantizations published2234
Smallest quantization8.39 GiB6.07 GiB
Q4_K_M15.66 GiB17.11 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.8-27BLlama3.2-30B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-UncensoredRatio
4,0960.25 GiB0.73 GiB2.94×
8,1920.50 GiB1.47 GiB2.94×
16,3841.00 GiB2.94 GiB2.94×
32,7682.00 GiB5.88 GiB2.94×
65,5364.00 GiB11.75 GiB2.94×
131,0728.00 GiB23.50 GiB2.94×
Qwen3.8-27B vs Llama3.2-30B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-Uncensored — size, memory and hardware fit — ossmodeldb