Skip to main content

​
The Instances page

gpuoutlet.ai/app/rentals is your control panel for every pod, current and historical. Three KPI cards at the top:
  • Active instances — currently Running
  • Hourly burn — combined hourly rate of every Running pod
  • Session spend — total cents charged across this billing session
Below: a list of pod cards filterable by status (All / Running / In progress / Stopped / Failed).

​
Stopping

Click any pod card → Stop button in the bottom-right.
Or via API
curl -X POST https://gpuoutlet.ai/v1/rentals/rent_01H.../stop \
  --cookie "sid=$SID"
Stop is:
  • Immediate — terminate signal sent within 100ms
  • Permanent — the pod is destroyed, its disk wiped
  • Settled — your wallet is debited for the exact seconds consumed, rounded to nearest second
There is no “pause” — pause-the-machine isn’t a meaningful concept for GPU rentals (the GPU is the expensive resource, not the disk). If you need to “pause” your work, save state to S3 / HuggingFace, stop the pod, and relaunch a new one later.

​
Restarting / relaunching

A stopped pod can’t be restarted. To pick up where you left off:
  1. Push your work to persistent storage before stopping
  2. Launch a fresh pod
  3. Pull your work back down
This sounds inconvenient but it’s actually a feature — every pod starts clean, so there’s no “my Python env is broken from last week” problem.

​
Reading the metering data

Each pod card shows:
  • Rate — $X.XX/hr
  • GPUs — e.g. 1× RTX 4090
  • Ran for — total runtime
  • Total — total charged
To get the second-level data: 
curl https://gpuoutlet.ai/v1/rentals/rent_01H.../meter \
  --cookie "sid=$SID" | jq
Sample
{
  "rentalId": "rent_01H...",
  "ratePerHour": 2.80,
  "secondsBilled": 4827,
  "currentChargeCents": 375,
  "lastHeartbeatAt": "2026-06-07T08:42:11Z"
}
secondsBilled / 3600 * ratePerHour * 100 should match currentChargeCents to the cent. We round half-up at settlement.

​
Filtering instance history

The status filter chips above the grid let you scope. Common queries:
  • Stopped — see your last week’s runs and how much each cost
  • Failed — see anything that didn’t provision; click in for the provider error log + auto-refund line

​
Auto-stop on launch (coming)

The Create flow will soon expose Auto-stop after N hours so you can guarantee a runaway script can’t burn through your wallet overnight. For now, set your wallet balance to the budget you’re comfortable with — every pod stops automatically when the wallet hits $0.

​
Bulk operations via API

If you’re running an experiment fleet:
Stop everything that's running
curl https://gpuoutlet.ai/v1/rentals?status=running --cookie "sid=$SID" | \
  jq -r '.rentals[].id' | \
  xargs -I{} curl -X POST https://gpuoutlet.ai/v1/rentals/{}/stop --cookie "sid=$SID"
Be careful with that command — there’s no undo.