Welcome to Prior Labs
TabPFN is a foundation model for tabular data - built to replace manual model tuning with instant, state-of-the-art predictions. It’s trained across millions of synthetic datasets to learn the learning process itself. Instead of optimizing weights for every new dataset, TabPFN already encodes the inductive biases, priors, and optimization strategies that traditional models must rediscover each time. When you run it on your data, it performs zero-shot inference - producing predictions in seconds that match or exceed the accuracy of tuned ensembles trained for hours.Why teams choose TabPFN
Accurate predictions in seconds
TabPFN-2.5 delivers benchmark-level performance instantly - achieving tuned-ensemble accuracy in seconds, not hours.
No training required
Skip the training loop. TabPFN performs zero-shot inference with no tuning or gradient descent.
Seamless integration
Plug into any workflow with the familiar
scikit-learn interface or through the Prior Labs API for production use.Robust in the real world
Handles missing values, outliers, and categorical features natively for stable, reliable performance.
Minimal preprocessing
Works directly on raw tabular data - no encoding, scaling, or imputation pipelines required.
Interpretable & trustworthy
Produces calibrated probabilities and integrates with SHAP for transparent, explainable outcomes.
Capabilities
Customers use TabPFN to power high-impact predictions - from disease diagnosis and risk modeling to energy forecasting and industrial optimization - across three core capabilities: classification, regression and time series forecasting.Classification
Predict categorical outcomes with calibrated probabilities and strong zero-shot accuracy.
Regression
Estimate continuous values with uncertainty-aware outputs and minimal preprocessing.
Forecasting
Model tabular time series (via TabPFN for forecasting) to predict future values and trends.