Nvidia made the full Run:ai platform available as open-source software within its DSX AI Factory platform on 13 October 2026, according to a company announcement. The release gives teams access to the integrated software for managing AI jobs and GPU capacity, as well as four components they can adopt separately. Nvidia continues to offer commercially supported versions.
Key points
- The release covers the integrated Run:ai platform and four independently adoptable components.
- Nvidia says the source code is on GitHub, while an installable Community Product is available through NGC.
- The Community Product needs no paid Run:ai subscription, but Nvidia enterprise support is not included.
Run:ai offers a stack and four components
Run:ai manages AI workloads running on Kubernetes, including data processing, training and inference. Nvidia describes its role as allocating GPU resources dynamically across those jobs. Rather than assigning computing capacity permanently to one task, dynamic allocation treats it as something shared: much as a workbench can be assigned to different jobs, GPU capacity can be assigned as workloads run. Nvidia says this approach can improve throughput and GPU utilisation on shared infrastructure.
Getting an answer to a routine question could depend on computing capacity shared with data processing and training. If those workloads ran on the same infrastructure, dynamic allocation would affect how capacity could be assigned while the answer was being generated.
The four separately adoptable parts are KAI Scheduler, GPU Fractioning Operator, KAI Resource Management and Workload Map. Nvidia says those components, together with additional platform code, make up the integrated Run:ai platform. That gives a team already operating its own platform a different route from installing Run:ai in full: it could incorporate an individual component where it needs that capability.
Nvidia also says partners can build products and services on the integrated platform or incorporate selected components. The distinction matters for companies assembling AI infrastructure from several systems, because the release offers both a complete orchestration stack and parts intended for use within another offering. Nvidia presents both routes as ways to make its GPU and workload-management software available across the DSX ecosystem.
GitHub code and the NGC Community Product
Nvidia says the complete Run:ai source code is in its Run:ai repository within the DSX AI Factory GitHub organisation. Teams can inspect the code, contribute to it, build on the platform or take its modular components separately. The source code is governed by the applicable open-source licences, according to Nvidia.
There is also an installation route for the integrated stack. Nvidia makes it available through NGC as a Community Product, with a Helm chart for installation. This separates access to the code from the packaged product: downloading the Community Product from NGC requires acceptance of the applicable Nvidia Software License Agreement and Product-Specific Terms for Nvidia AI Products.
Nvidia says the Community Product requires neither a paid Run:ai subscription nor a Nvidia support entitlement. It does not include Nvidia enterprise support. A team choosing between the community download and a supported commercial offering would therefore be choosing a support arrangement as well as a route to the software.
DSX partners take different routes to Run:ai
Nvidia names eight partners integrating or offering Run:ai, with uses that span managed Kubernetes, AI clouds and inference services. It says Crusoe integrates the software with Crusoe Managed Kubernetes, while Mirantis extends k0rdent AI with it. Nebius offers the complete platform as a one-click application deployment in Nebius AI Cloud, and OpenNebula Systems integrates it with the OpenNebula AI Cloud Platform through OneKS.
The other arrangements also differ. Nvidia says Saturn Cloud integrates Run:ai into its Token Factory platform for inference services, Spectro Cloud integrates it into PaletteAI, and vCluster provides it through deployment templates. Rafay is expanding its Run:ai integration to offer it as a fully managed service, according to Nvidia. That expansion is described as an undertaking, rather than one of the existing integrations in the company’s list.
For enterprise AI factory deployments, Nvidia says commercially supported Run:ai offerings remain available as a standalone product, through Nvidia Mission Control and Nvidia AI Enterprise, and in select cloud marketplaces. Availability, licensing and support terms vary by offering.
