gpu

Use runpodctl to list available GPU types, include unavailable options, and review hardware specifications from the command line.

List available GPU types and their specifications.

Command

runpodctl gpu <subcommand> [flags]

Subcommands

List available GPUs

List GPUs that are currently available. A GPU counts as available when it has stock in Secure Cloud or Community Cloud:

runpodctl gpu list

Include unavailable GPUs in the list:

runpodctl gpu list --include-unavailable

List flags

--include-unavailable · bool

Include GPUs that are currently unavailable.

Example output

[
  {
    "available": true,
    "communityCloud": true,
    "displayName": "RTX 4090",
    "gpuId": "NVIDIA GeForce RTX 4090",
    "memoryInGb": 24,
    "secureCloud": true,
    "stockStatus": "High"
  },
  {
    "available": true,
    "communityCloud": true,
    "displayName": "A100 PCIe",
    "gpuId": "NVIDIA A100 80GB PCIe",
    "memoryInGb": 80,
    "secureCloud": true,
    "stockStatus": "High"
  }
]

Using GPU IDs

When creating Pods or Serverless endpoints, use the GPU ID from the list with the --gpu-id flag:

runpodctl pod create --template-id runpod-torch-v21 --gpu-id "NVIDIA RTX 4090"

Related commands

• runpodctl pod create
• runpodctl serverless create
• runpodctl datacenter list