Start Here
Welcome to Grafilab
Grafilab gives you two ways to get work onto a GPU: rent a whole machine, or call a model through an API. Both live behind one account, one prepaid wallet, and one console at https://app.grafilab.ai.
What Grafilab is
Section titled “What Grafilab is”Grafilab is two products sharing an account:
- Compute rents you an on-demand GPU machine from a marketplace of supplier nodes. You pick the hardware, pick a prebuilt template, pay for a fixed duration up front, and get root over SSH plus a public web URL for whatever the template runs.
- Models is an inference API in front of a hosted model catalog — text, embeddings, images, and video. It speaks both the OpenAI and the Anthropic wire formats, so most tools work after you change a base URL and a key.
One prepaid balance pays for both a GPU rental and an API call, so you never top up in two places — see Credits & Billing.
Which half do you need
Section titled “Which half do you need”A shell, a notebook, or a diffusion UI, on hardware you manage yourself. Pick a GPU and a template, prepay at least 12 hours, then connect over SSH or the instance’s public URL.
I want an APIA model endpoint instead of a server to look after. Generate a key, point your OpenAI- or Anthropic-compatible client at Grafilab, and call a model id from the live catalog. Billed per token.
Nothing stops you using both — a rented box for experiments and the API for production traffic is a common split.
The console at a glance
Section titled “The console at a glance”Everything you do outside your own code happens in the console at app.grafilab.ai. Each sidebar entry below links straight to that page.
| Sidebar entry | What it’s for |
|---|---|
| Home | Dashboard: balance, recent activity, shortcuts. |
| Cloud → Compute | The three-step deploy wizard for a new GPU instance. |
| Cloud → My Instance | Your running instances, with Connect, Web Access, Renew, and Backup. |
| Models → Playground | Try a model in the browser before you write any code. |
| Models → API | Generate, rename, and revoke API keys; see per-key usage. |
| My Settings | Your profile, including your country. See Settings & Profile. |
| My Settings → Team | Invite people to act on your account and choose what they may do. See Team & Permissions. |
| Billing | Available credits, monthly usage breakdown, invoice history. |
The header shows Current Balance and a Buy Credit button on every page, so topping up never takes you out of what you were doing. Beside them is the account picker, which is how you act on someone else’s account when they have invited you — see switching accounts. Which entries you see depends on your permissions.
Docker Templates (/docker-template) has no sidebar entry at all; you reach
it by URL or from the deploy wizard.
The Home dashboard
Section titled “The Home dashboard”Home is the landing page after sign-in, titled Dashboard. Three panels matter:
- Inference Token Cost — a spend chart over your API usage, toggled between Last 7 Days and Last 30 Days.
- Active Instances — live monitoring of your running GPUs, reading “No active instances tracking.” until you deploy one.
- Quick Deploy — three shortcut cards: GPU Instance, Inference API, and Docker Template.