Runpod Adds Docker and Network Volume Support to Enhanced CPU Pods
Runpod says its Enhanced CPU Pods now support Docker and network volumes, according to the title of a post on the company blog. The same page points to updated PyTorch templates, an explanation of Global Volumes and a guide to Serverless billing.
- Runpod's post title says Enhanced CPU Pods now support Docker and network volumes.
- PyTorch templates now cover 2.6 through 2.13 across multiple CUDA and Ubuntu versions, per Runpod.
- A new CUDA dropdown in the Pod deploy flow helps match templates to GPU and workload.
- Runpod says Global Volumes mount like a network volume but run on object storage.
- A separate post explains why Serverless balances can drop when jobs fail to complete.
Enhanced CPU Pods gain Docker and network volumes
The headline of the Runpod post is "Enhanced CPU Pods Now Support Docker and Network Volumes." It signals two additions for CPU-based Pods on the platform: the ability to work with Docker and the ability to attach network volumes.
The text available from the page does not describe the CPU Pod change itself. It gives no details on how Docker support works, which CPU configurations are covered, or when and where the feature is available. No pricing is mentioned either. Those points should be checked in the full announcement before anyone plans a deployment around them.
Updated PyTorch templates and a CUDA dropdown
The page also highlights work on Runpod's PyTorch templates. According to the company, the templates "now cover 2.6 through 2.13 across multiple CUDA and Ubuntu versions." Runpod has also added a new CUDA dropdown to the Pod deploy flow.
The stated purpose of the dropdown is to let users "match a template to your GPU and workload." For teams that run PyTorch jobs on Pods, this should make it easier to pick an environment that fits the hardware they have chosen, instead of working around a mismatched default.
Global Volumes and Serverless billing explained
Two other posts are teased alongside the announcement. The first is about Global Volumes, which Runpod says are not "network volumes, but better." The company describes them this way: "Global Volumes mount like a network volume but run on object storage." Runpod adds that this "changes what you should put on them," which suggests the two storage types suit different kinds of data.
The second post covers how Serverless billing works when a job fails to complete. According to the teaser, it explains why a balance can drop even when nothing succeeds, how to find where the money went, and how to prevent it from happening again.
What it means for users
For users, the clearest takeaways are practical. CPU Pods are described as gaining Docker and network volume support, PyTorch users get a wider range of template versions plus a CUDA selector, and storage choices are being clarified with the distinction between network volumes and Global Volumes. Anyone running Serverless workloads has a dedicated explanation of failed-job charges.
The source text stops short of specifics on availability, limits or costs, so the full Runpod posts are the place to confirm details.
RunPod: Enhanced CPU Pods Now Support Docker and Network Volumes
