Multi-node is finally here, plus a fresh new look for the entire platform.
[NEW]
-
Multi-node clusters with 3.2Tbps InfiniBand are now available. Spin up distributed training jobs across multiple H100/H200 nodes. Your existing storage volumes work out of the box. Attach a volume and it's available across all nodes in your cluster. More documentation coming next week.
-
Major UI overhaul across the entire platform. We've redesigned the compute, storage, and settings pages to be cleaner and easier to use. Expect UI improvements with every update going forward.
[COMING NEXT]
-
Cloudy CLI to manage everything from your terminal. Works with Claude Code too, so you can spin up GPUs without leaving your editor.
-
Templates to get you started quickly. Large scale distributed training, finetuning, or simple single node experiments.
-
Detach and attach volumes on live pods. Swap volumes without stopping your instance.
-
Significant capacity expansion. We'll have the most available H100/H200 cloud capacity. No more waiting for GPUs.
See you next Friday :)