# soda-inria/tabicl

TabICLv2: An open tabular foundation model

Repository: https://github.com/soda-inria/tabicl
Canonical: https://ross.abutalabs.com/products/tabicl
Homepage: https://tabicl.readthedocs.io/en/latest/
Language: Python
License: NOASSERTION
License Family: other
Topics: deep-learning, foundation-models, machine-learning, tabular-data, tabular-methods, tabular-model
Last push: 2026-08-26T16:22:59+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 81, longevity 40
- inputs: {"age_days": 572, "days_push": 7, "days_rel": 126, "gap_med": 7.0, "n_releases_24m": 17}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1309, forks 175 (observed 2026-08-28T04:04:19.432759+00:00)

## What it is
TabICLv2 is an open-source tabular foundation model that performs classification and regression via a single forward pass through a pre-trained transformer, using in-context learning instead of training. It is pip-installable, scikit-learn compliant, and achieves state-of-the-art benchmark results without hyperparameter tuning.

## Use cases
- classify tabular data without training a model
- predict on tabular datasets without hyperparameter tuning
- replace XGBoost or CatBoost with a faster tabular model
- run zero-shot regression on structured data
- fit and predict tabular data on a GPU in seconds
- scale tabular inference to hundreds of thousands of samples

## When to choose
- you need state-of-the-art tabular classification or regression with no tuning
- you want fast inference on medium-sized tabular datasets with a GPU
- you prefer a scikit-learn-compatible API
- you want an open-source, permissively licensed tabular foundation model

## When to avoid
- you have no GPU and very large datasets
- you need strict interpretability out of the box
- you work with non-tabular data like images or text
- you require guaranteed accuracy on datasets far beyond 500K samples

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, transformers
- domain: machine-learning, data-science, large-language-models
- platform: python, cross-platform
- tags: tabular-data, foundation-model, in-context-learning, classification, regression, scikit-learn, tabular-deep-learning, gpu

## Member repositories
- soda-inria/tabicl (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:19.432759+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:50:00.907983+00:00, confidence not recorded.
  - readme: https://github.com/soda-inria/tabicl (fetched 2026-08-28T04:04:19.432759+00:00, sha e42119543adf)
- Data as of 2026-08-30T08:39:29.467469+00:00.