About
What is this?
A reference site tracking tabular foundation models: the pre-trained models, the companies building them, and the research behind them. Updated as the field evolves.
Who is this for?
- Investors evaluating the tabular AI market
- ML engineers and data scientists comparing models for production use
- Researchers following the state of the art in tabular pre-training
What are tabular foundation models?
Pre-trained models that make predictions on structured data (tables, spreadsheets, databases) without traditional task-specific training. You pass in a new dataset and get predictions in a single forward pass. No hyperparameter search, no gradient descent at inference time.
Most tabular foundation models are trained on large collections of synthetic or real-world tabular datasets, then generalize to unseen tabular tasks (classification, regression, time series) at test time. The leading approaches use transformer architectures with in-context learning.
How to cite this website
Molnar, Christoph. “Tabular Foundation Overview.” 2026. https://tabular-foundation.christophmolnar.com.
BibTeX:
@misc{molnar2026tabularfoundation,
author = {Christoph Molnar},
title = {Tabular Foundation Overview},
year = {2026},
url = {https://tabular-foundation.christophmolnar.com}
}Contact
Have a correction or suggestion? Reach out at tfm@christophmolnar.com.