Model comparison

Llama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-Uncensored vs Qwen3.6-27B

At Q4_K_M, Llama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-Uncensored is the smaller download — 11,091,316,864 bytes against 16,547,397,888. At long context the gap widens: Qwen3.6-27B's KV cache at 32K is 1.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Llama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-UncensoredQwen3.6-27B
Parameters18.0B27.8B
Architecturellamaqwen35
Layers2864
Native context131,072262,144
Mixture of expertsyes, 8 expertsno
Quantizations published3540
Smallest quantization3.72 GiB8.74 GiB
Q4_K_M10.33 GiB15.41 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextLlama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-UncensoredQwen3.6-27BRatio
4,0960.44 GiB0.25 GiB1.75×
8,1920.88 GiB0.50 GiB1.75×
16,3841.75 GiB1.00 GiB1.75×
32,7683.50 GiB2.00 GiB1.75×
65,5367.00 GiB4.00 GiB1.75×
131,07214.00 GiB8.00 GiB1.75×