Skip to main content
Select a model family for details. See Model licensing for license information.

​
TabPFN-3.5 family

TabPFN-3.5 builds on TabPFN-3 and improves most on non-IID settings, where rows are not independent, features are not clean and data types are mixed. It excels at predictions on tables that are text-rich, have high-cardinality and high-dimensionality features, as well as grouped and temporal data. The family includes four versions of the models: *The hosted TabPFN-3.5-Plus API also has text preprocessing limits: at most 50 million tokens in total and 2,500 characters per text value. For TabPFN-3.5, we recommend datasets with up to 6,000 columns; the model ceiling is 20,000 columns. Row and column limits trade off, so their maximum values cannot necessarily be used together. The hosted API allows up to 500 million training-upload cells and 200 million training cells per estimator context. Uploads above 200 million cells require row subsampling. Use the many-class extension for classification tasks with more than 160 classes. The model is licensed under TABPFN-3-5 License v1.0. Code under Apache 2.0.

​
TabPFN-3 family

TabPFN-3 supports numerical and categorical data, missing values, and up to 160 classes natively. The family includes a local base model, Plus for native text features, and Thinking for when compute matters less than accuracy. Row and column limits trade off, so their maximum values cannot necessarily be used together. Use the many-class extension for classification tasks with more than 160 classes. The model is licensed under TABPFN-3 License v1.0. Code under Apache 2.0.

​
TabPFN-2.6

TabPFN-2.6 was the default model in TabPFN OSS package versions v7.0.0–v7.1.x. It shares the same architecture and dataset constraints as TabPFN-2.5 with improved benchmark performance. The model is licensed under TABPFN-2.6 License v1.0. Code under Apache 2.0.

​
TabPFN-2.5 family

TabPFN-2.5 expanded support to larger datasets than TabPFNv2. The family includes a base model for numerical and categorical data and a hosted Plus model that also handles text. Read the TabPFN-2.5 model report. The model weights are licensed under TABPFN-2.5 License v1.1. Code under Apache 2.0.

​
TabPFNv2

TabPFNv2 is the model published in Nature and the first TabPFN to handle real-world messy data — mixed types, missing values, and categorical features automatically. The model weights are licensed under the Prior Labs License (Apache 2.0 with an additional attribution requirement): classifier license, regressor license. Code under Apache 2.0.

Nature Publication

Read about TabPFNv2 in our Nature publication from 2025.

​
TabPFN Model License

The code in the TabPFN repository is licensed under Apache 2.0, starting with tabpfn 9.0.0. Model weights are licensed separately: TabPFN-2.5, TabPFN-2.6, TabPFN-3 and TabPFN-3.5 weights under non-commercial licenses, and TabPFNv2 weights under the Prior Labs License. Each model section above links its license.
No. Accepting the license unlocks the model weights download. Your datasets stay local and are never sent to Prior Labs, and the tabpfn package works offline once the weights are downloaded. See Offline use.
Enterprises can download and experiment with the model internally without cost. This includes benchmarking, testing on internal datasets, exploring capabilities, and understanding fit for use cases. Any competitive benchmarking (i.e. for procurement), testing in deployment in production environments, use in workflows that support business decisions, client projects, or revenue generating activities requires a separate commercial license or API agreement.
Enterprises may run preliminary internal assessments on their own data for research and exploration. This includes testing on internal datasets, exploring capabilities and understanding fit for use-cases.
You cannot use the model or its outputs for any activity that influences business decisions, evaluation of vendors, procurement decisions, or commercial workflows. This includes proof of value that affects internal budgeting or vendor selection. In those cases, a commercial license is required. For that, check out our pricing or contact our sales team.
Not without a commercial license. Any evaluation that influences commercial decisions or internal production planning is considered commercial use. For that, check out our pricing or contact our sales team.
You can run small scale internal experiments. If you intend to integrate TabPFN into a product or platform, the API or a commercial license is required. For that, check out our pricing or contact our sales team.
You will need a commercial license or API agreement. Check out our pricing or contact our sales team.
Yes. The same restrictions apply to all derivatives. Fine tuned, modified, or otherwise adapted models remain non-commercial unless you secure a commercial license.
Outputs are yours, but they can only be used for non-commercial evaluation and research. Using them in production or for any business decisions requires a commercial license.
If you do not use any TabPFN model weights, checkpoints, or other licensed model elements, then the non-commercial license does not apply to that separately developed model. However, if you use TabPFN (or its outputs) to train, guide, fine tune, or distill another model for commercial use, that is not permitted under this license.
No. Using TabPFN (or its outputs) to train, fine tune, distill, or otherwise improve a commercial model is not allowed without a commercial agreement.
Use our API or obtain a commercial license. The API is the simplest path for production trials and product integration. For that, check out our pricing or contact our sales team.
Prior versions remain available under their original licenses. TabPFN-2.5, TabPFN-2.6, TabPFN-3 and TabPFN-3.5 have non-commercial model licenses.
Please contact our sales team or check out our pricing.