Model comparison

LTX-2 vs LTX2.3-10Eros

At Q4_K_M, LTX-2 is the smaller download — 12,840,118,752 bytes against 14,296,160,672.

From the file· summed bytes, KV per layer

Side by side

LTX-2LTX2.3-10Eros
Parameters18.9B21.0B
Architectureltxvltxv
Layers
Native context
Mixture of expertsnono
Quantizations published2836
Smallest quantization7.48 GiB9.63 GiB
Q4_K_M11.96 GiB13.31 GiB
Licence

KV cache by context

the term that decides long-context viability
ContextLTX-2LTX2.3-10ErosRatio
4,096
8,192
16,384
32,768
65,536
131,072