Prior Labs

Overview

Prior Labs is a German AI research company focused on tabular foundation models: machine-learning systems designed for spreadsheets, databases, and other structured data. Its main technology, TabPFN, applies a pretrained model to prediction tasks without requiring a separately trained model for every dataset.

The company is notable for combining academic research, open-source distribution, and enterprise software development in a segment that has received less attention than language and image models. Its work targets applications including risk assessment, forecasting, churn analysis, and scientific or medical data analysis.

History

Prior Labs was founded in late 2024 by Frank Hutter, Noah Hollmann, and Sauraj Gambhir. The company grew from research on TabPFN, a foundation model for tabular data, and raised €9 million in equity financing in 2025. TabPFN was published in Nature in 2025 and became the basis for the company’s open-source and commercial work. SAP announced an acquisition in 2026, with Prior Labs continuing as an independent legal entity and retaining its brand, team, and research commitments.

Product Lines

Manufacturing

Prior Labs does not manufacture physical products. It develops AI models and software in-house, with TabPFN distributed through open-source releases and commercial interfaces or services.

Notable Products

• TabPFN - Foundation model for prediction and classification on structured tabular datasets, available in open-source and commercial forms.
• TabPFN-2.5 - Newer TabPFN generation designed to handle larger tabular datasets and more features than earlier versions.
• TabPFN API - Application programming interface for integrating Prior Labs’ tabular-data analysis capabilities into business workflows.

Reputation

Prior Labs is regarded by machine-learning researchers and data-science professionals as a serious specialist in tabular foundation models, supported by the academic background of its founders and advisers. TabPFN has attracted substantial open-source use, citations, and attention after its Nature publication. Its strengths include pretrained performance, reduced need for dataset-specific training, and accessibility to researchers. Limitations include uncertainty around performance across all enterprise datasets, scaling constraints relative to general-purpose models, and limited publicly disclosed pricing and production details.