Introducing TabFM: A zero-shot foundation model for tabular data

TabFM is a foundation model designed for tabular data classification and regression that eliminates the need for manual model training, hyperparameter tuning, and complex feature engineering by framing tabular prediction as an in-context learning problem. Instead of traditional training for each new task, TabFM treats the entire dataset—including historical training examples and target testing rows—as a single unified prompt.

Source: Google Research | 2026-06-30