LLM Reference

TabFM Models by Google

GoogleNoncommercialOpen weights
1 model2026

Last refreshed 2026-07-01. Next refresh: weekly.

Details

ResearcherGoogle
Commercial useCommercial use: non-commercial
Models1
Released2026

About

TabFM is Google's tabular foundation model for zero-shot classification and regression on structured tables. It frames tabular prediction as an in-context learning problem, using training examples and target rows as a unified prompt so users can generate predictions in a single forward pass without per-dataset model training, hyperparameter tuning, or manual feature engineering. Google Research released TabFM with public GitHub code and downloadable pretrained weights on Hugging Face, and says BigQuery integration via AI.PREDICT is planned.

Decision facts

Best fit
tabularcodingstructured outputs
Capability starting point
TabFM 1.0.0
Lowest tracked input
Not tracked
Closest related family
Google Cloud Text-to-Speech

Current Variants

Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.

1 in view

Use when the workload needs tabular.

2026-06tabular

Release Timeline

1 release group
2026-06
1 current
Current

Specifications(1 models)

TabFM model specifications comparison
ModelReleased
TabFM 1.0.02026-06