- dstack provides ML teams with a unified control plane for GPU provisioning and orchestration - across cloud, Kubernetes, and on-prem. It streamlines development, training, and inference — reducing costs 3–7x and - preventing lock-in. + dstack is a GPU-native orchestration built for modern AI teams. It simplifies + GPU provisioning and workload management across clouds, Kubernetes, and on-prem — + through a single unified control plane.
- Managing AI infrastructure requires efficient GPU orchestration, whether workloads run - on a single GPU cloud, across multiple GPU providers, or on-prem clusters. + Managing AI infrastructure requires efficient GPU orchestration tightly integrated with open-source training and + inference frameworks.
- dstack provides an open stack for GPU orchestration that streamlines development, training, - and inference, and can be used with any hardware, open-source tools, and frameworks. + dstack provides a unified control plane—so workloads stay portable, reproducible, and infrastructure remains interchangeable.
@@ -173,22 +172,21 @@- dstack natively integrates with top GPU clouds—automating cluster provisioning and - workload orchestration to maximize efficiency and minimize overhead. + dstack natively integrates with leading GPU clouds for fast, efficient provisioning.
- It can provision and manage VM clusters through native integrations or via Kubernetes. + It can provision and manage GPU VMs directly through cloud APIs, or operate through Kubernetes when required.
- + Backends -
- For provisioned Kubernetes clusters, connect them to dstack using the Kubernetes backend. - If you run vanilla bare-metal servers or VMs without Kubernetes, use SSH fleets - instead. + If you already run Kubernetes on-prem, connect your cluster to dstack using the Kubernetes backend. +
+ ++ For bare-metal servers or VMs without Kubernetes, use SSH fleets to orchestrate GPUs directly.
- + SSH fleets - - + Kubernetes -
- -- Either way, connecting existing on-prem clusters to dstack takes just minutes. -
- Before training or deployment, ML engineers explore and debug their code. + Before training or deploying models, ML engineers need interactive GPU access to experiment and debug.
-dstack's dev environments make it easy to connect your - desktop IDE to powerful cloud or on-prem GPUs—streamlining the entire development loop. +
dstack's dev environments let you connect desktop IDEs such as VS + Code, Cursor, and + Windsurf directly to cloud or on-prem GPUs.
+ class="md-button md-button-secondary"> Dev environments - +
- Move from single-instance experiments to multi-node distributed training without friction. dstack lets you define complex jobs with a simple configuration, - handling the scheduling and orchestration for you. + Run training or batch workloads on a single GPU, or scale to multi-GPU and multi-node clusters using simple task configurations. + dstack automates cluster provisioning, resource allocation, and job scheduling.
-This allows your team to focus on research while ensuring that expensive cluster resources are utilized efficiently.
++ During execution, dstack reports GPU utilization, memory usage, and GPU health metrics for each job. +
- With dstack, you can easily deploy any model as a secure, - auto-scaling OpenAI-compatible endpoint, all while using your custom code, Docker image, and - serving framework. + With dstack, you can deploy models as secure, + auto-scaling, OpenAI-compatible endpoints, integrating with top open-source serving frameworks. +
+ ++ dstack supports disaggregated prefill/decode and cache-aware routing, providing + production-grade, optimized inference.