Product Overview

Two compute modes, one operational experience.

Choose between self-serve GPU instances and reserved multi-GPU clusters for predictable training and production operations.

Why teams choose this path

Capabilities built for high-throughput ML operations

Clear path between self-serve instances and reserved clusters
Technical specs, launch workflows, and pricing context on every product surface
Consistent CTA journey for both developer-led and procurement-led buying motions

What you can do

Launch notebooks, SSH sessions, and batch runs with visible queue states and workload lifecycle tracking.

Plan multi-node capacity for recurring training windows with priority scheduling and commitment-based pricing.

Coordinate security, procurement, and migration planning with a dedicated solutions and support motion.

Self-serve GPUs for notebooks, training jobs, and debugging

Reserved multi-node capacity with priority scheduling