TabPFN

TabPFN is a pre trained tabular foundation model from Prior Labs that makes predictions on structured data such as spreadsheets and database tables. Instead of training a new model for every dataset, a user fits it to their data and gets predictions in a single forward pass, with no hyperparameter tuning and no feature engineering pipeline. It handles classification and regression, and it reads free text, columns of thousands of distinct identifiers and rows with hundreds of measurements directly as features. The current model family is TabPFN 3.5, which includes a Thinking mode for harder prediction tasks. The tool is aimed at data scientists and machine learning engineers who want a strong prediction baseline quickly. It is used through a Python SDK that follows the scikit learn pattern, a REST API and an MCP server. The open source tabpfn package runs the model weights on local hardware, and the vendor recommends a GPU for anything beyond small datasets. Deployment options are the hosted API, which processes data in the EU, a private cloud deployment on platforms such as AWS SageMaker and Azure AI Foundry, and an on premises container for air gapped environments. The hosted API has a free plan with daily and monthly usage limits that includes the latest model and Thinking mode. Pro and Max plans add higher limits, extended caching, dedicated support and enterprise security options, and both are sold through the sales team without published prices. The open weights package and the private cloud tier are free for non commercial use, while production use under a commercial licence is also arranged through sales.