TabFM
Google ResearchTabFM is a tabular foundation model from Google Research that classifies or regresses on a new table without a training step: the user's labeled rows are read as context and predictions for new rows come from a single forward pass. It handles mixed numeric and categorical columns and is pretrained only on synthetic tables. The tabfm package wraps it in scikit-learn-style estimators with JAX and PyTorch backends.
Openness
2 high confidence- weights
- open (1.0.0 and 1.1.0 checkpoints in PyTorch and JAX on the Hub, ungated; the package downloads them on first use)
- data
- closed (synthetic tables drawn from a structural causal model prior; the generator is not released)
- code
- partial (inference and estimator code; no pretraining code)
- license
- Apache-2.0 (code)+TabFM-Non-Commercial-License-v1.0 (weights; non-commercial, non-production use only)
The tabfm code is Apache-2.0, but the pretrained weights the package downloads by default are under Google's TabFM Non-Commercial License, which allows testing, evaluation and research only and bars production use. The synthetic training data cannot be regenerated, since neither the generator nor the pretraining code is released.
- https://huggingface.co/api/models?author=google&search=tabfm&limit=100&expand[]=downloads&expand[]=cardData&expand[]=gated recorded 2026-11-12
google/tabfm-1.0.0-pytorch, tabfm-1.1.0-pytorch and tabfm-1.0.0-jax, all gated false, license_name tabfm-non-commercial-v1.0.
- https://huggingface.co/google/tabfm-1.1.0-pytorch/raw/main/LICENSE recorded 2026-11-12
TabFM Non-Commercial License v1.0: grants use "solely for your Non-Commercial Purposes"; use "for any revenue-generating activity" or "in direct or indirect interactions with end users or production systems" is not a Non-Commercial Purpose.
- https://huggingface.co/google/tabfm-1.1.0-pytorch/raw/main/README.md recorded 2026-11-12
tabfm-1.1.0-pytorch card: "TabFM was trained on hundreds of millions of synthetic datasets generated dynamically using structural causal models (SCMs)"; weights under the TabFM Non-Commercial License v1.0, source code Apache 2.0.
- https://raw.githubusercontent.com/google-research/tabfm/main/LICENSE recorded 2026-11-12
LICENSE is the Apache License, Version 2.0, January 2004.
- https://raw.githubusercontent.com/google-research/tabfm/main/README.md recorded 2026-11-12
README: "The TabFM source code in this repository is licensed under Apache-2.0. However, the default Quick Start calls tabfm_v1_0_0.load(), which automatically downloads pretrained weights from Hugging Face. Those pretrained weights are distributed under the separate tabfm-non-commercial-v1.0 license and are restricted to non-commercial, non-production use."
- https://ungh.cc/repos/google-research/tabfm/files/main recorded 2026-11-12
Repository tree of 49 paths: JAX and PyTorch model, checkpoint loading, estimator, Hub conversion and example code; no training or prior-generation script.
Adoption
2 medium confidenceHugging Face downloads over the trailing 30 days for the three TabFM checkpoint repositories, most of them for the 1.0.0 PyTorch weights. Monthly PyPI downloads of the tabfm package, which fetches those weights, fall in the same range.
- https://huggingface.co/api/models?author=google&search=tabfm&limit=100&expand[]=downloads&expand[]=cardData&expand[]=gated recorded 2026-11-12
tabfm-1.0.0-pytorch 38,604, tabfm-1.0.0-jax 2,073 and tabfm-1.1.0-pytorch 981 downloads in the trailing 30 days (41,658 together).
- https://pypistats.org/api/packages/tabfm/recent recorded 2026-11-12
last_month 22,340 downloads of tabfm
Capability
4 medium confidenceTabFM reads the user's training rows as context and predicts on new rows without fitting or tuning, which is what TabPFN does. It is one model, limited to ten classes, and does not search over or ensemble other learners.
- https://huggingface.co/google/tabfm-1.1.0-pytorch/raw/main/README.md recorded 2026-11-12
Card: "requiring no fine-tuning or hyperparameter search - training examples are passed as context and predictions are made in a single forward pass"; "Max 10 classes"; "evaluated on TabArena across 51 datasets (38 classification, 13 regression)"; a TabFMClassifier.ensemble() preset.
- https://raw.githubusercontent.com/google-research/tabfm/main/README.md recorded 2026-11-12
README: "At inference time, TabFM does not require training parameters on your dataset; instead, it leverages in-context learning by reading your training data as "context""; TabFMClassifier and TabFMRegressor examples.
Verified 2026-11-12