Comment by woofwoofwoof

TabPFN [1] is also working in this area and getting better and better. The inference time is still much worse than any of the more traditional models (xg/catboost etc).

[1] https://github.com/PriorLabs/tabpfn