Model comparison

Qwen3.8-27B vs writing-roleplay-20k-context-nemo-12b-v1.0

At Q4_K_M, writing-roleplay-20k-context-nemo-12b-v1.0 is the smaller download — 7,477,218,048 bytes against 16,810,714,336. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 2.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.8-27Bwriting-roleplay-20k-context-nemo-12b-v1.0
Parameters27.8B12.2B
Architectureqwen35llama
Layers6440
Native context262,1441,024,000
Mixture of expertsnono
Quantizations published2224
Smallest quantization8.39 GiB3.85 GiB
Q4_K_M15.66 GiB6.96 GiB
Licenceapache-2.0—

KV cache by context

the term that decides long-context viability
ContextQwen3.8-27Bwriting-roleplay-20k-context-nemo-12b-v1.0Ratio
4,0960.25 GiB0.63 GiB2.50×
8,1920.50 GiB1.25 GiB2.50×
16,3841.00 GiB2.50 GiB2.50×
32,7682.00 GiB5.00 GiB2.50×
65,5364.00 GiB10.00 GiB2.50×
131,0728.00 GiB20.00 GiB2.50×