Dataset Handling
Supports datasets up to 50,000 samples and 2,000 features (TabPFN-2.5), with larger models handling up to 10 million rows, automatically managing missing values and categorical data.
Multi-Task Support
Enables classification (binary and multi-class with calibrated probabilities), regression with uncertainty estimates, time-series forecasting, anomaly detection, data generation, fine-tuning, interpretability, and text integration within tables.
Deployment Options
Available via hosted API for commercial use and as an open-source Python package on Hugging Face for non-commercial use, with scikit-learn compatible interface and PyTorch/CUDA support.
Integration
Integrates with Python notebooks, production pipelines, enterprise platforms, on-premises environments, private clouds, and Google Sheets.
Performance
Delivers predictions in seconds without tuning or retraining, outperforming baseline methods like ridge regression and gradient boosting in speed and accuracy on tasks such as crop yield forecasting.