Platform Overview

What you manage in ANK

At its core, ANK is about instances. You pick a tier, choose a configuration, pick a base image, give it access credentials, and manage the workload from there.

Core primitives

• Instance tier — the hardware class. Currently available: NVIDIA GeForce RTX 3090 (S, M, 2X), NVIDIA GeForce RTX 5080 (S, M, L), and NVIDIA GeForce RTX 5090 (S, M, XL, XXL, 2X). Coming Soon: RTX PRO 6000 (S, M, L, XL) and DGX Spark (4XL).
• Configuration — the CPU, RAM, disk, and GPU allocation within a tier (S / M / L / XL / XXL / 2X / 4XL).
• Node layout — whether a configuration lives on one host or spans several. 2X tiers are two independent instances on two hosts; everything else is a single instance on one host. See Single-Node Instances and Multi-Node Instances.
• Billing plan — pay-as-you-go (hourly), daily, weekly, or (for Pro users) monthly.
• Image — the software the instance boots from: Ubuntu CPU, CUDA, or PyTorch.
• Access credentials — a username and password, SSH public keys, or both.

Lifecycle

create → provision → active → stop or delete

Keep reading with Pricing and Billing and Availability before you commit to a long run.