gpu environments

List launch environments (certified frameworks, versions, and raw-VM OS options)

Synopsis

Lists the server-controlled launch environment catalog — every value the --environment flag on gpu instances create accepts:

• Frameworks — the certified framework roster (certified:, e.g. certified:pytorch), each with a one-line "what's included" summary. The bare certified base image (drivers + CUDA) is always available.
• Versions — where a framework publishes more than one build for a CUDA generation, the version picker options (certified:@, e.g. certified:pytorch@2.13). Versions are tagged with the CUDA generation they target — a version only launches on GPUs that map to it.
• Operating systems — the raw-VM OS catalog (raw-vm:, e.g. raw-vm:ubuntu-24.04), with whether the NVIDIA driver is preinstalled or installed at boot.

The catalog is global; whether a specific offering can serve a given framework/version/OS is enforced at launch (the server returns a 422 with a human-readable detail if not). Public — no auth required.

gpu environments [flags]

Examples

# List frameworks, versions, and OS options
gpu environments

# JSON for scripting
gpu environments --output json | jq '.frameworks'

# Then launch with a chosen environment
gpu instances create -t h100_sxm --environment certified:pytorch@2.13
gpu instances create -t h100_sxm --environment raw-vm:ubuntu-24.04

Options

  -h, --help   help for environments

Options inherited from parent commands

      --api-base string   API base URL (override with GPUAI_API_BASE env) (default "https://api.gpu.ai/v1")
      --debug             Enable debug logging to stderr
  -o, --output string     Output format: table|json (default table on TTY, json otherwise)

SEE ALSO

• gpu - GPU.ai command-line interface
• gpu instances create - Launch a GPU instance (accepts --environment)