Loading tool…
Open-source control plane that provisions and schedules GPU jobs across clouds, Kubernetes and on-prem.
dstack is a control plane for GPU work. You declare fleets, dev environments, training tasks, inference services and volumes in config files, and it handles provisioning, job queuing, autoscaling and networking across whatever compute you point it at: cloud GPU providers, a Kubernetes cluster or on-prem racks, covering NVIDIA, AMD, TPU and Tenstorrent hardware. The core is open source under the Mozilla Public License 2.0 and free to self-host. dstack Inc. also runs dstack Sky, a hosted version with access to a GPU marketplace, and an enterprise build adding SSO and air-gapped installation. Neither hosted tier publishes a price.
The open-source control plane is free under MPL 2.0 and self-hostable. dstack Sky and the enterprise self-hosted edition are not priced on the site: Sky bills GPU marketplace usage and the enterprise route ends at a discovery call.
Use tool ↗dstack's pitch is not owning GPUs but not caring whose they are, so the same config runs against a cloud marketplace, your Kubernetes cluster or a rack in the office. The MPL licence means you can run the whole control plane yourself and never speak to the vendor. It is a developer tool through and through: config files and a CLI, no dashboard-first experience, and no published prices for the hosted tiers.
Watch out: Self-hosting is free, but the GPU bill still lands wherever the compute actually runs. dstack schedules the work, it does not discount it.
No reviews yet — be the first to review Dstack.
Reviews are tied to your EffectHub account — one review per tool, so the rating for Dstack reflects real users.
No questions yet — ask the first one about Dstack.
Questions about Dstack are tied to your EffectHub account, so answers can reach you and the section stays free of spam.