Model comparison

Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-Thinking vs Qwen3.8-27B

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 23,936,565,984. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-ThinkingQwen3.8-27B
Parameters39.5B27.8B
Architectureqwen35qwen35
Layers9664
Native context262,144262,144
Mixture of expertsnono
Quantizations published3422
Smallest quantization9.38 GiB8.39 GiB
Q4_K_M22.29 GiB15.66 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-ThinkingQwen3.8-27BRatio
4,0960.38 GiB0.25 GiB1.50×
8,1920.75 GiB0.50 GiB1.50×
16,3841.50 GiB1.00 GiB1.50×
32,7683.00 GiB2.00 GiB1.50×
65,5366.00 GiB4.00 GiB1.50×
131,07212.00 GiB8.00 GiB1.50×