There are two main differences between the available subscriptions: the maximum size of your Datasets* and the number of tasks that can be performed in parallel.
|Max. Task Size
The subscription properties decide the kind of servers deployed in the auto-scaling infrastructure managed by BigML which flexibly adapts to ensure that most of the tasks will start immediately by dynamically minimizing queuing time.
*Larger files can be uploaded to BigML. They generate a Source resource that can be used to sample them into a smaller Dataset that can fit the subscription limits.