Skip to content

swm gpus

Search live GPU availability and pricing across all configured providers.

Terminal window
swm gpus [OPTIONS]
OptionDescription
-g, --gpuGPU name filter (fuzzy match)
-c, --countMinimum GPU count
--max-priceMaximum price per hour
-p, --providerFilter to specific provider
--secureOnly show secure cloud offers
-r, --regionFilter by region (free text, e.g. us-west)
--sortSort by: price (default), vram, provider
-n, --limitMax results to show
--allShow all results (no limit)
Terminal window
swm gpus -g h100 --max-price 3.00 --sort price
swm gpus -g a100 -p runpod --secure
swm gpus -g "rtx 4090" -c 4
swm gpus -r us-west
swm gpus --all --sort vram

The result table includes a Min CUDA column showing each GPU’s minimum CUDA toolkit (e.g. Hopper → 11.8, Ada → 11.8, Blackwell → 12.8). When all rows in the filtered result share the same minimum, swm appends --cuda <X.Y> to the suggested next-step swm pod create command:

Next → swm pod create -p <provider> -g h200 -n <name> --cuda 11.8

Pair this with swm images list -p <provider> --cuda <X.Y> to see compatible Docker images, or just pass --cuda to pod create and let swm pick the newest matching image automatically. See CUDA & image compatibility for the full picture.

GPU names are matched with a 3-pass strategy: exact match, substring match, then token-based match. If no match is found, swm suggests similar names.