>>107096817
Omni models just need to be a lot bigger to compensate for retardation.
We are basically feeding them more actual new data rather than refining old one.