# Prior Labs

> Prior Labs builds Tabular Foundation Models (TFMs) — pre-trained AI models that make state-of-the-art predictions on structured (tabular) data. Their flagship product, TabPFN, achieves best-in-class results on classification and regression tasks without task-specific training, feature engineering, or ML pipelines. Prior Labs entered a definitive agreement to be acquired by SAP in May 2026, with a planned €1B+ investment over four years to scale it into a globally leading frontier AI lab for structured data.

Prior Labs was founded by Frank Hutter, Noah Hollmann, and Sauraj Gambhir, and is headquartered in Freiburg, Germany, with offices in Berlin and New York City. The company's open-source TabPFN model has been published in Nature and has surpassed 3 million downloads. Scientific advisors include Yann LeCun and Bernhard Schölkopf.

## Products

- [TabPFN-3](https://priorlabs.ai/tabpfn): The latest flagship tabular foundation model. Achieves 93% win rate over classic ML on TabArena benchmark. Handles up to 1M rows with 0.2s inference. Includes a Thinking mode (TabPFN-3-Plus) that beats AutoML in 80% of cases (+420 ELO). Supports classification, regression, and all standard tabular tasks without tuning. Works with raw data including missing values, outliers, and categoricals.
- [TabPFN-v2](https://priorlabs.ai/tabpfn-2): Previous generation tabular foundation model, still available.
- [TabPFN-TS](https://priorlabs.ai/tabpfn): Time-series variant of TabPFN for forecasting tasks.

## Deployment

- [Deployment Overview](https://priorlabs.ai/deployment): TabPFN can be deployed via managed API, private cloud (AWS SageMaker, Azure AI Foundry, Google Vertex AI), self-hosted (download weights from Hugging Face), Databricks Lakehouse integration, Snowflake, and MCP (Model Context Protocol) for AI agents.
- [Model Context Protocol](https://priorlabs.ai/deployment/model-context-protocol): Embed TabPFN into AI agents via MCP for structured data intelligence in agentic workflows.
- [Databricks](https://priorlabs.ai/deployment/databricks): Deploy TabPFN inside a Databricks Lakehouse.
- [AWS SageMaker](https://aws.amazon.com/marketplace/pp/prodview-chfhncrdzlb3s): One-click deployment in AWS cloud (external).
- [Azure AI Foundry Model Catalog](https://priorlabs.ai/deployment/azure-ai-foundry-model-catalog): Deploy TabPFN in Microsoft Azure infrastructure.

## Industries

- [Finance](https://priorlabs.ai/industries/finance): Credit risk, fraud detection, algorithmic trading, mortgage approvals, financial forecasting.
- [Healthcare](https://priorlabs.ai/industries/healthcare): Cancer risk prediction, genetic variant classification, Alzheimer's disease prediction, clinical decision support.
- [Industrials](https://priorlabs.ai/industries/industrials): Predictive maintenance, defect detection, pipeline compliance classification.
- [Energy](https://priorlabs.ai/industries/energy): Energy trading forecasting, meter data classification, grid anomaly detection.
- [Tech](https://priorlabs.ai/industries/tech): Churn prediction, media spend forecasting, synthetic data generation.

## Case Studies

- [Taktile — Financial Risk Management](https://priorlabs.ai/case-studies/taktile): TabPFN for fraud detection with minimal labeled training data.
- [BostonGene — Immune System Profiles](https://priorlabs.ai/case-studies/boston-gene): 90% accuracy in cancer patient identification from peripheral blood.
- [Hitachi Rail — Predictive Maintenance](https://priorlabs.ai/case-studies/hitachi): 40% improvement in anomaly detection for rail track maintenance.
- [Oxford Cancer Analytics (OxCan) — Liquid Biopsy](https://priorlabs.ai/case-studies/oxcan): Actionable clinical insights from complex proteomic datasets for lung disease.
- [Creditplus Bank — Car Loan Approvals](https://priorlabs.ai/case-studies/credit-plus): Credit decisioning in motor financing using TabPFN.
- [TD Bank — Financial Forecasting](https://priorlabs.ai/case-studies/td-bank): Enterprise-scale financial prediction with Layer 6 AI.
- [Exito — Media Spend Forecasting](https://priorlabs.ai/case-studies/exito): Replaced legacy models with more accurate, less manual spend forecasting.
- [NHS / UHNM — Clinical Intubation Prediction](https://priorlabs.ai/case-studies/nhs): 22 of 24 outcomes correctly predicted for critical care decisions.

## Research

- [Research Overview](https://priorlabs.ai/research): Prior Labs publishes research across reasoning, causality, time-series forecasting, scalability, interpretability, and fairness in tabular AI.
- Key research areas: multi-table relational reasoning, causal inference from observational data, zero-shot time-series forecasting, scaling in-context learning to millions of rows, interpretable predictions for high-stakes domains, and counterfactual fairness.
- Notable papers: DoPFN (causal effect estimation via in-context learning), FairPFN (counterfactual fairness with transformers), Drift-Resilient TabPFN (temporal distribution shifts). TabPFN-2 was published in Nature (January 2025).

## Developer Resources

- [Documentation](https://docs.priorlabs.ai/): Full docs including overview, API reference, and agentic MCP setup.
- [API Reference](https://docs.priorlabs.ai/api-reference/getting-started): REST API and Python SDK for accessing TabPFN.
- [GitHub](https://github.com/PriorLabs/): Open-source repositories including TabPFN, TabPFN-TS, and more.
- [Hugging Face](https://huggingface.co/Prior-Labs): Model weights and hub page.
- [Playground](https://ux.priorlabs.ai/playground): Try TabPFN in the browser without any setup.
- [Discord](https://discord.com/invite/VJRuU3bSxt): Community support and discussion.

## Company

- [About](https://priorlabs.ai/about): Company mission, team, investors, and scientific advisory board. Founded by Frank Hutter, Noah Hollmann, Sauraj Gambhir, and Bernhard Schölkopf. Investors include XTX Ventures and others; angel investors include Thomas Wolf (Hugging Face), Robin Rombach (Black Forest Labs), and Ed Grefenstette (DeepMind).
- [Blog](https://priorlabs.ai/blog): News, product updates, and technical writing from the team.
- [Technical Reports](https://priorlabs.ai/technical-reports): In-depth model reports, including the TabPFN-3 report.
- [Careers](https://priorlabs.ai/careers): Open positions at Prior Labs.
- [Manifesto](https://priorlabs.ai/blog-posts/manifesto): Vision for Multimodal Tabular Foundation Models as the foundation for agentic AI systems that reason over tables, language, and images.
- [SAP Acquisition Announcement](https://priorlabs.ai/blog-posts/priorlabs-next-chapter): Prior Labs' next chapter under SAP with €1B+ investment.

## Legal

- [Privacy Policy](https://priorlabs.ai/privacy-policy)
- [Terms of Service](https://priorlabs.ai/general-terms-and-conditions)
- [Acceptable Use Policy](https://priorlabs.ai/aup)
- [Imprint](https://priorlabs.ai/imprint)