Xiaomi-TabLDM uses one pretrained model across tabular datasets, performing classification and regression without retraining or task-specific tuning.

Xiaomi released Xiaomi-TabLDM as a general-purpose foundation model for tabular, or structured, data. The announcement was made through Xiaomi Technology (小米技术).

The system uses a single pretrained model and a unified default configuration across different tabular datasets. Xiaomi said users do not need to retrain or separately tune the model for every task.

Xiaomi-TabLDM supports both classification and regression predictions without task-specific retraining, parameter tuning, or post-processing ensembles. The source did not provide benchmark scores, training-data volume, parameter count, supported table sizes, inference-hardware requirements, or licensing terms.

The announcement concerned the model’s release and stated operating approach. It did not disclose production deployments, customer adoption, commercial pricing, or independently validated comparisons with other tabular-learning systems.