NVIDIA releases Kumo Relational, a no-training predictor for databases
NVIDIA has released Kumo Relational, a foundation model that predicts directly on relational databases without training and outperforms supervised models across 41 tasks.
By GennnAi··1 min read
Tested on 41 tasks across 15 databases, it beats supervised relational deep learning models and expert-engineered gradient-boosted tree pipelines. On the SAP SALT benchmark, it surpasses AutoGluon ensembles by 8% in mean reciprocal rank. The model scales to databases with more than 500 billion rows. It is available as open source through NVIDIA libraries or as an NIM, and agents can query it through a predictive query language.
Source: NVIDIA, published on October 21, 2026.