# Nori

Canonical page: https://www.synthefy.com/product/tabular

Nori is an open-weight foundation model for tabular prediction. It returns predictions without feature engineering, training, or hyperparameter tuning.

Nori can run locally on private data. Its code and model weights are licensed under Apache 2.0 and are free for commercial use.

## Quickstart

```python
from synthefy_nori import NoriRegressor

model = NoriRegressor()
predictions = model.predict(
    X_test, X_train, y_train
)
```

Install the package with `pip install synthefy-nori`. The weights download on first use and are cached locally. Data preparation and inference compute are still required; "no training" refers to avoiding task-specific model fitting.

## What is Nori?

Nori is an open-weight foundation model for tabular prediction.

## Does Nori require training or hyperparameter tuning?

No. Pass test rows with labeled context in one `predict()` call, and Nori returns predictions without fitting a new model.

## Can Nori run locally on private data?

Yes. Install `synthefy-nori`, download the weights once, and run Nori locally. The weights are cached after the first download.

## Is Nori free for commercial use?

Yes. Nori's code and model weights are licensed under Apache 2.0 and are free for commercial use.

## Links

- [Documentation](https://docs.synthefy.com/nori/)
- [Python quickstart](https://docs.synthefy.com/nori/quickstart)
- [Managed API](https://docs.synthefy.com/nori/quickstart#api): hosted Nori predictions as an alternative to running the Python package locally.
- [Source code](https://github.com/Synthefy/synthefy-nori)
- [Model weights](https://huggingface.co/Synthefy/Nori)