The only model that I found so far that extract table data with OCR is dots.ocr. Models that came after it have not done a good job. Interesting on testing this new model.