Batch Jobs

What it does

Batch Jobs takes your container, runs it to completion, and stores the output. If a peer goes offline mid-run, your job is retried elsewhere. You don't manage servers, queues, or retry logic — you submit a job and check back when it's done. Ideal for video encoding, scientific computing, dataset preprocessing.

Unlike a persistent compute pod, you only pay for time actually spent computing. Failed peer attempts don't bill you. Jobs can be parallelized across many peers when the work is splittable, which means a 10-hour batch can finish in 30 minutes if you size it right.