Compute
Deploy a GPU Instance
By the end of this page you have a running GPU container of your own: a root shell over SSH and, if you picked a web template, a public URL. The console walks you through it in three steps and one checkout screen.
Prerequisites
Section titled “Prerequisites”- A Grafilab account with a country set on your profile. The country stamps your wallet currency, and rentals are refused without it. See Create Your Account.
- Enough credits to cover the whole rental, or a card you can pay with at checkout. See Credits & Billing.
- An idea of which template you want. Browse Templates first if you’re not sure — the wizard doesn’t let you change it after launch.
Deploy the instance
Section titled “Deploy the instance”-
In the console, open Compute.
A sticky stepper reads Hardware Selection, Template Selection, Review & Launch. It advances on its own as you make each selection, and you can scroll back up to change an earlier choice.
-
Under Choose Your Power, click the card for the GPU you want.
The subtitle reads “Pick the compute hardware — GPU or CPU — that fits your workload. We’ll match you with the best available node.” Each card shows the family name, a positioning line, and two spec bullets. The row is built from live supplier inventory, so only listed families appear; a Contact Us ribbon means sales-arranged, not self-serve.
-
Under Template Selection — “Select a pre-configured software environment.” — click the template card you want.
Your own templates appear first, with a My Template badge and a ⋮ menu offering Edit and Delete.
-
Under Review & Launch, choose a node and click Deploy.
Nodes are listed cheapest first, three at a time; Load More reveals three more. Each row shows the GPU name, a CPU model badge, a reputation badge (
excellentorgood), a DC Verified badge for data-centre hosts,VRAMandRAM, the location, and the price with a /hr unit. More expands the spec grid — vCPU, RAM, VRAM, Storage.If the list is empty you’ll see “No nodes available for NVIDIA L4 right now. Try another GPU.” — go back to step 2 and pick a different family.
-
On Review & Deploy Instance, pick a rental period and a payment method, then click Launch Instance.
The left column repeats the node with its Base Rate per hour. Rental Period Selection offers duration cards — the period label, “12 hours continuous”, and the total. Payment Method lists what is available to you; the credit option shows Balance: so you can see whether the wallet covers it.
The sticky Order Summary on the right confirms Instance, Region, OS Image, Rental Duration, then Subtotal and Total Charge. Below the button, the legal line spells out the commitment: “By launching, you agree to the Service Terms and will be charged MYR 1.45/hour, for a total of MYR 17.43.”
-
Wait for the payment page to confirm, then click My Instance.
Paying from credit takes you straight there. Paying by card sends you to Stripe checkout first and returns you afterwards. Success reads “Your Payment is Successful” with “Thank you for your payment. Your compute agent will be ready in few minutes.”; failure reads “Your Payment is Unsuccessful” with a Back to Rent and Run button.
If deployment itself fails you’ll get a Deployment failed toast — “Failed to deploy. Please try again.” Nothing is running, so start again from step 1.
Verify
Section titled “Verify”Open My Instance. Within a few minutes the card for your rental should show:
- Status Active.
- Footer lines Running grafilab/jupyterlab:2026-06-09 with the template’s image, and Expire at 9 Sept 2026, 04:32 am MYT with the end of your rental.
- An IP Address, core count, and speed for the node.
- A working Connect button: it opens the Connect with SSH dialog with a copyable SSH Connect: command and Password:.
Paste that command into a terminal, enter the password, and you land in a root shell. If SSH refuses right after the card appears, wait a minute and retry — the container’s network link finishes coming up just after the card does.
What you’re charged
Section titled “What you’re charged”The whole rental period is charged at launch, in one transaction, at the rate on the checkout page — there is no per-minute metering and no early-stop refund. Nothing else is billed while the instance runs; see Credits & Billing for when each product takes the money.
At Expire at the container is removed with everything written inside it. Back up or Renew before then — both are covered in Manage Instances.

