Model comparison

Qwen3.8-27B vs L3-DARKEST-PLANET-16.5B

At Q4_K_M, L3-DARKEST-PLANET-16.5B is the smaller download — 10,207,765,536 bytes against 16,810,714,336. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 4.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3.8-27BL3-DARKEST-PLANET-16.5B
Parameters27.8B16.5B
Architectureqwen35llama
Layers6471
Native context262,1448,192
Mixture of expertsnono
Quantizations published2210
Smallest quantization8.39 GiB7.00 GiB
Q4_K_M15.66 GiB9.51 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.8-27BL3-DARKEST-PLANET-16.5BRatio
4,0960.25 GiB1.11 GiB4.43×
8,1920.50 GiB2.22 GiB4.44×
16,3841.00 GiB4.44 GiB4.44×
32,7682.00 GiB8.88 GiB4.44×
65,5364.00 GiB17.75 GiB4.44×
131,0728.00 GiB35.50 GiB4.44×