Model comparison

Wan_2.1_ComfyUI_repackaged vs LTX2.3-10Eros

At Q4_K_M, LTX2.3-10Eros is the smaller download — 14,296,160,672 bytes against 68,351,109,664.

From the file· summed bytes, KV per layer

Side by side

Wan_2.1_ComfyUI_repackagedLTX2.3-10Eros
Parameters21.0B
Architecturepigltxv
Layers
Native context
Mixture of expertsnono
Quantizations published4836
Smallest quantization4.25 GiB9.63 GiB
Q4_K_M63.66 GiB13.31 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextWan_2.1_ComfyUI_repackagedLTX2.3-10ErosRatio
4,096
8,192
16,384
32,768
65,536
131,072