Anyscale

Per-minute billing for compute is the most concrete thing about Anyscale, and the rates vary a lot by hardware.

CPU-only capacity starts at $0.00006 per minute when you deploy in your own cloud.

GPU options in your cloud run from NVIDIA T4 at $0.00246 per minute up to A100 80GB at $0.02941 per minute, with AWS Trainium1 and Inferentia2 also available.

Anyscale is an AI application platform for building, running, and scaling AI applications, and it sits in the AI developer tools category.

You can also run workloads in Anyscale's own cloud, but the same hardware costs more there.

CPU-only starts at $0.00855 per minute, a T4 at $0.01643, an A10G at $0.02723, and an A100 80G at $0.11312.

If you already have cloud accounts, deploying in your own looks like the cheaper route.

Anyscale has a freemium model, so you can start without paying, though you should check how far the free usage goes before planning a larger project.