Get Started

Getting Started with InstantGPU

URL: /docs/getting-started/

Get your first AI workload running on GPU infrastructure in a few simple steps.

1 Create Account

→

2 Choose GPU

→

3 Configure

→

4 Deploy

→

5 Connect & Run

Create Your Account

Create an InstantGPU account to access your dashboard, GPU infrastructure, deployments, and AI services. Once registered, generate your API Key from your dashboard to enable programmatic deployments and SDK access.

Create InstantGPU Account →

Choose Your Workflow

Choose the infrastructure option that best matches your workload:

GPU Cloud

Use a dedicated GPU when you need full control over the machine, persistent workloads, or a custom environment.

Serverless GPU

Use serverless infrastructure when your workload needs automatic scaling or variable GPU capacity with scale-to-zero.

AI Endpoints

Use ready-to-integrate AI APIs when you want to add AI capabilities without deploying and managing the underlying model yourself.

Choose a GPU

Select your GPU based on:

Model size: Weights and quantization parameters (e.g. 8B, 70B, 405B).
VRAM requirements: Minimum memory footprint needed in GB (e.g. 24GB, 48GB, 80GB, 141GB).
Performance & Architecture: Ada Lovelace, Hopper, or Blackwell tensor cores.
Expected traffic: Requests per second and concurrent stream counts.
Latency requirements: Time to first token (TTFT) and throughput needs.
Budget: Low-cost spot/community vs enterprise secure cloud SLAs.

Deploy

Choose a pre-built template or configure your own environment:

Pre-built AI Templates

Deploy ready-to-use Docker images for PyTorch, vLLM, ComfyUI, Ollama, Whisper, and TensorFlow in 1 click.

Custom Container Images

Bring your own Docker image from Docker Hub or private registries with your proprietary code and weights.

Connect

Depending on your deployment, connect using:

SSH

Direct terminal access to your dedicated GPU instance via secure key.

REST API

HTTP POST/GET calls with JSON payloads and Bearer token auth.

Endpoint URL

Direct public or private HTTPS endpoints with automated load balancing.

Application SDK

Native Python, TypeScript, and Node.js SDK libraries.

Build, deploy, and scale your AI workloads with InstantGPU.

Ready to deploy your first container or query an inference endpoint?

Explore Serverless Architecture →