Google Research has proposed a new foundation model called TabFM that treats tabular prediction as an in-context learning problem, allowing it to generate predictions for a new, unseen table in a single forward pass. For enterprise developers and AI engineers, this reduces time-to-production from weeks of pipeline engineering to a single API call. Traditional machine learning requires training a new model from scratch for every dataset, with hyperparameter tuning loops, feature engineering, and retraining pipelines to fight data drift. Weihao Kong, Research Scientist at Google Research, noted that traditional models "incur ongoing operational debt through data drift monitoring and retraining pipelines to stay accurate." Large language models (LLMs) struggle with tables due to context limits, tokenization inefficiency that destroys numerical precision, and structural blindness when a 2D table is serialized as a 1D text string. To run inference with TabFM, you do not update any model weights; instead, you take historical examples and target rows and pass them as a single, unified prompt.