This is usually due to the dataset size and the amount of fields in it. The bigger your dataset is, the longer BigML will take to build the model. Nevertheless, BigML can build models from many thousands of instances within just a few seconds. Larger datasets can operate in streaming mode. Keep in mind that speed is also dependent on your subscription plan or specific Private Deployment (Virtual Private Cloud or On-Premises) as more memory, processors and parallel tasks can greatly improve performance.