Prior Labs completes TabPFN-4 training for tabular predictions

priorlabs.ai6 min read

Prior Labs announced the completion of training for TabPFN-4, its next-generation tabular foundation model featuring a sub-quadratic architecture that scales to 10 million rows. According to the company, the model outperforms previous versions and key baselines across multiple benchmarks, including large, temporal, and grouped data regimes as well as Kaggle competitions. Open article →