Get started

Welcome

Deploy dedicated GPUs in minutes and pay by the second.

Featherless GPU Cloud runs AI and general compute workloads on dedicated GPU and CPU compute. Deploy an instance for development or a long-running service, submit a Run Job for a one-off command, and keep data on network volumes that outlive both.

Get started

Quickstart

Deploy your first instance, run code on it, and stop it when you're done.

Concepts

The terms used across the console, the CLI and these docs.

Credentials

Add SSH keys, create API keys and connect private registries.

Install the CLI

Manage instances, Run Jobs and files from your terminal.

Products

Instances

Dedicated GPU or CPU compute for development, notebooks and services.

Run Jobs

Run a command on temporary compute and pay only while it runs.

Storage

Keep datasets, models and checkpoints on network volumes.

Templates

Start from an official environment or save your own launch settings.

Examples

Develop in JupyterLab

Open JupyterLab on a GPU instance with your work on a network volume.

Serve an LLM with vLLM

Run an OpenAI-compatible model server on a public HTTPS address.

Fine-tune with a Run Job

Train a LoRA adapter on temporary compute and keep it on a network volume.

Launch GPUs from a CPU jumpbox

Work from a small CPU instance and start GPUs only when you need them.

Support

Discord

Ask questions, report problems and talk to the team.