Skip to main content
An instance (or “pod” — we use the terms interchangeably) is a single rental — a VM with one or more GPUs attached, running the template image you picked, with your SSH key injected at first boot.

Lifecycle states

StateWhat’s happeningBilling
ProvisioningPod allocated, image pulling, SSH config writingNot billed
RunningSSH reachable, work happensMetered per second
StoppingTerminate request sent to providerFinal partial second settled
StoppedPod released, IP unboundNo further charges
FailedAllocation or boot errorRefunded automatically

Templates

A template is just a starting OS image plus pre-installed software. Six are wired in today:

Ubuntu 24.04

Plain. CUDA drivers, that’s it. Default pick.

PyTorch

Ubuntu + CUDA + PyTorch 2.x + transformers + accelerate.

vLLM

Inference server preloaded. Pass a model with --model org/name.

Ollama

Ollama daemon on port 11434, ready to ollama pull.

A1111 (Stable Diffusion)

AUTOMATIC1111 WebUI on port 7860.

text-generation-webui

Oobabooga’s text-gen on port 7860.
Custom images are coming. The settings page already has a placeholder for it — see Settings.

Per-second metering

Every running pod sends a heartbeat every second. The metering worker:
  1. Reads the pod’s hourly rate
  2. Converts to per-second: rate_per_hour / 3600
  3. Atomically debits that amount from the rental’s wallet ledger entry
  4. If the wallet reaches $0, sends a terminate signal to the provider
You can verify the math at any time:
gpuout rental show rent_01H... --metered
# {"seconds_billed": 4827, "rate_per_hour": 2.80, "current_charge": 3.753}

Stopping

Click Stop in the dashboard or hit POST /v1/rentals/:id/stop. The pod is unreachable within ~5 seconds. We settle the final partial second so you pay exactly for what you used, not a rounded-up minute.
Stopping is permanent. The pod’s local disk is wiped. Save your work to external storage (S3, your machine via scp, a HuggingFace repo) BEFORE hitting Stop.

Failure & refund

If a pod fails to provision or crashes within 60 seconds of going Running, the wallet is refunded in full automatically. We log the provider-side error and you’ll see it in the rental detail view.