Model comparison

Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP vs Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness

At Q4_K_M, Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness is the smaller download — 5,463,652,192 bytes against 36,545,830,912. At long context the gap widens: Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness's KV cache at 32K is 4.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness
Parameters27.8B9.1B
Architectureqwen35qwen35
Layers6416
Native context262,144262,144
Mixture of expertsnono
Quantizations published342
Smallest quantization10.12 GiB4.88 GiB
Q4_K_M34.04 GiB5.09 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-MadnessRatio
4,0960.25 GiB0.06 GiB4.00×
8,1920.50 GiB0.13 GiB4.00×
16,3841.00 GiB0.25 GiB4.00×
32,7682.00 GiB0.50 GiB4.00×
65,5364.00 GiB1.00 GiB4.00×
131,0728.00 GiB2.00 GiB4.00×
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP vs Qwen3.6-9B-Heretic-Uncensored-Thinking-Sweet-Madness — size, memory and hardware fit — ossmodeldb