Standard GPU Pods

Cost-effective GPU rental from a distributed provider network. The right choice when you need raw compute at the best price and hardware confidentiality is not a requirement.

Standard vs Confidential, which one?

Deploy a Standard GPU pod

1. Open Standard GPUs and filter by GPU, price, location or VRAM.
2. Pick a machine, price shown is your final per-hour price, billed per second.
3. Choose a Docker template (PyTorch, CUDA, Ubuntu...), or your own public image through the API (below).
4. Select or add an SSH key, required for access.
5. Deploy. 1 hour is prepaid at deploy; if you stop earlier, the unused time is refunded per second.

Deploy with the API

Standard pods deploy with an API key too, through the same deployment as the dashboard (same price, same prepaid hour). List the machines, then deploy one by its machine_id:

curl -H "X-API-Key: $VOLT_API_KEY" \
  "https://api.voltagegpu.com/api/volt/machines?provider=standard"

curl -X POST https://api.voltagegpu.com/api/volt/pods \
  -H "X-API-Key: $VOLT_API_KEY" -H "Content-Type: application/json" \
  -d '{"provider": "standard", "name": "my-pod", "machine_id": "<machine_id>"}'

To run your own public image instead of a template, add image, and optionally one simple start_command, the ports to expose and env variables. An image that cannot be prepared answers 400 and nothing is charged:

curl -X POST https://api.voltagegpu.com/api/volt/pods \
  -H "X-API-Key: $VOLT_API_KEY" -H "Content-Type: application/json" \
  -d '{"provider": "standard", "name": "vllm-server", "machine_id": "<machine_id>",
       "image": "vllm/vllm-openai:latest",
       "start_command": "vllm serve Qwen/Qwen2.5-7B-Instruct --port 8000",
       "ports": [8000], "env": {"HF_TOKEN": "hf_xxx"}}'

Then poll GET /api/volt/pods/<id> until ssh_command is set, and release with DELETE /api/volt/pods/<id>. Step by step: Create a pod, with the API. Full reference: API reference, Standard Pods.

Billing

• Per-second billing with 1 hour prepaid at deployment.
• Stopping a pod within the first hour refunds the unused prepaid fraction automatically, from the dashboard or with DELETE /api/volt/pods/<id>.
• A pod that the provider never manages to start gets the full prepaid hour back.
• Every stopped pod, with what it actually cost, is listed under Pod history on Your Pods.
• A machine is rented whole: the hourly price covers all of its GPUs.
• Pods are stopped automatically if your balance reaches zero.

Check the machine you got

Standard machines come from many hosts, so check yours in one minute. This read-only script compares the machine with its listing: GPU model and count, VRAM, RAM, disk, and a 25 MB download speed test (skip it with --no-net). Nothing about the machine is sent anywhere. It exits 0 when everything matches and 2 when something does not:

curl -fsSL https://voltagegpu.com/hwcheck.sh | bash -s -- --gpu "RTX 4090" --gpus 1 --vram 24 --ram 128 --disk 500

The pod page gives you this command with your listing's values filled in. If the machine is not what was listed, release it within the first hour: the unused time is refunded.

SSH access

SSH keys are managed centrally in your dashboard and work for both Standard and Confidential pods. Once the pod is running, the connection command appears on the pod detail page.