The original article brags about speed but you can see by the four tables I added in the article that the execution speed is slowest for TabPFN. Given the fact they don't recommend using it for dataset with more than 10K rows, the speed was not an issue for me