Post #2068037

I think DeepSeek has proven (and continues to prove) that you don't need incredibly power-hungry models to get good results. Part of the inefficiency with the current models is likely just American Logic (TM) where bigger is better and high operating costs are a way to attract *more* capital, not less.