Welcome
CosmicAC
Managed GPU compute for machine learning and compute-intensive workloads. Get direct access to GPU hardware, run open-source language models, and scale compute workloads without managing infrastructure.
What CosmicAC offers
- Dedicated GPU container: A GPU Container job is an isolated virtual machine (VM) with dedicated GPU hardware, accessible through an interactive shell session.
- Language model serving: Managed Inference allows you to run open-source models without provisioning or managing serving infrastructure.
- Full environment control: GPU Container jobs give you full control over the VM environment. You can configure the operating system, install packages, and run custom scripts as needed.
- Fast job startup: CosmicAC provisions resources and schedules jobs automatically. Jobs start in minutes.
- Isolated, secure workloads: Every job is a KubeVirt VM with direct GPU access and hardware-level isolation.
- Real-time error alerts: CosmicAC sends email notifications when errors occur.
Get Started
Concepts
How-to guides
How to create a GPU Container job from the web UI
Create a GPU container job from the CosmicAC web interface.
How to access a GPU container
Connect to a running GPU container and open an interactive shell session.
How to connect to Managed Inference
Configure the CLI and send your first inference request.