Model comparison

Qwen3.8-27B vs Qwythos-9B-Claude-Mythos-5-1M-uncensored-heretic

At Q4_K_M, Qwythos-9B-Claude-Mythos-5-1M-uncensored-heretic is the smaller download — 6,216,967,936 bytes against 16,810,714,336. At long context the gap widens: Qwythos-9B-Claude-Mythos-5-1M-uncensored-heretic's KV cache at 32K is 2.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.8-27BQwythos-9B-Claude-Mythos-5-1M-uncensored-heretic
Parameters27.8B9.4B
Architectureqwen35qwen35
Layers6432
Native context262,1441,048,576
Mixture of expertsnono
Quantizations published2221
Smallest quantization8.39 GiB5.53 GiB
Q4_K_M15.66 GiB5.79 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.8-27BQwythos-9B-Claude-Mythos-5-1M-uncensored-hereticRatio
4,0960.25 GiB0.13 GiB2.00×
8,1920.50 GiB0.25 GiB2.00×
16,3841.00 GiB0.50 GiB2.00×
32,7682.00 GiB1.00 GiB2.00×
65,5364.00 GiB2.00 GiB2.00×
131,0728.00 GiB4.00 GiB2.00×