NVIDIA has made the full NVIDIA Run:ai platform available as open source software, giving customers, developers, and ecosystem partners access to production-grade AI workload and GPU orchestration capabilities. The release expands the NVIDIA DSX AI Factory Platform by establishing Run:ai as an open foundation for the broader ecosystem to build differentiated solutions.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
NVIDIA releases the complete Run:ai platform as open-source software, positioning it as the open foundation of the DSX AI Factory and splitting out four independently adoptable modules; eight ecosystem partners have begun integrating, while enterprise-grade support remains on the commercial product line.
The boundary of NVIDIA's move this time is clear: release the complete Run:ai platform as open-source software, use it to extend the NVIDIA DSX AI Factory platform, and position it as the open foundation for AI workload and GPU orchestration. This is not tossing a single-point tool to the community, but opening up the entire orchestration layer that previously existed as a commercial product.
The release simultaneously splits out four independently adoptable modular components: KAI Scheduler, responsible for GPU-aware Kubernetes scheduling; GPU Fractioning Operator, used to enable CUDA-native GPU sharing on Kubernetes clusters; KAI Resource Management, providing organizational quota and hierarchical management capabilities; and Workload Map, providing a consistent workload management contract. Together with the rest of the platform code, they constitute the complete Run:ai platform.
This dual-layer structure of "whole plus modules" determines the subsequent adoption path: platform teams and ecosystem partners can either use it as an integrated whole or take just one piece to shore up their own shortcomings.
Among the four modules, KAI Scheduler's degree of openness is the most explicit: it is the Kubernetes-native scheduler that drives the Run:ai platform, open-sourced under the Apache 2.0 license, and has been contributed to the CNCF as a Sandbox project.
More critical is the migration of the development path—KAI remains the upstream engine behind Run:ai, and all scheduler development takes place in the community repository, meaning subsequent improvements will flow through the same codebase rather than forking between an open-source version and a commercial version.
What needs to be distinguished: the source only clarifies KAI Scheduler's license and foundation affiliation; whether the other three components have also been contributed to the CNCF or other foundations is not stated by the source.
The complete source code is located in the NVIDIA Run:ai repository under the NVIDIA DSX AI Factory GitHub organization, where teams can view the code, participate in development, build on it, or independently adopt its modular components.
The complete platform is also available on NVIDIA NGC as a Community Product, accompanied by a Helm chart for installing the integrated stack. The barrier to this path is kept very low: no paid Run:ai subscription is required, nor is an NVIDIA support entitlement.
But "free" has a clear boundary. Downloading the Community Product from NGC requires accepting the applicable NVIDIA software license agreement and NVIDIA AI product-specific terms; the source code itself is governed by its applicable open-source license; and the Community Product does not include NVIDIA enterprise support.
Eight NVIDIA DSX ecosystem partners are integrating the open-source platform into their respective products: Crusoe, Mirantis, Nebius, OpenNebula Systems, Rafay, Saturn Cloud, Spectro Cloud, and vCluster.
The integration approaches given by the source vary widely: Crusoe integrates with Crusoe Managed Kubernetes; Mirantis extends k0rdent AI; Nebius offers one-click deployment on Nebius AI Cloud; OpenNebula integrates through OneKS; Rafay provides a fully managed service; Saturn Cloud supports white-label inference services through Token Factory; Spectro Cloud integrates into PaletteAI; and vCluster provides a pre-validated deployment template.
These details come from the source's account and are source claims rather than independently verified facts. Also unstated are the actual implementation timelines and complete functional scope of each partner's integration—that is, what can be confirmed at present is "is integrating," not "already available."
For NVIDIA Enterprise AI Factory deployments, Run:ai continues to be offered through NVIDIA-supported commercial products, including the standalone product, NVIDIA Mission Control, NVIDIA AI Enterprise, and some cloud marketplaces, with availability, licensing, and support terms varying by product.
This creates two parallel paths: the open-source side provides code and community collaboration, while the commercial side provides enterprise support. The source does not state the specific impact of this open-source release on existing paid Run:ai customers, such as migration paths or pricing changes.
It can be inferred from the source that NVIDIA, by open-sourcing Run:ai and joining forces with eight ecosystem partners, intends to build DSX AI Factory into the open foundation for AI infrastructure orchestration, while retaining enterprise-grade support as a differentiated product. This is an inference, not a conclusion directly stated by the source.
NVIDIA claims that after open-sourcing, every DSX ecosystem partner and AI cloud can increase workload throughput, improve GPU utilization, and operate shared AI infrastructure more efficiently.
This is a source claim, and it is not supported by quantitative data—the specific magnitude of improvement is not given in the source. What can be confirmed is only a directional statement, not a verifiable benefit figure.
In addition, the source mentions two related open-source projects, Model Streamer and Grove, which complement the platform by accelerating model loading and orchestrating multi-component inference workloads.
The specific date of the open-source release is October 13, 2026, but this date comes from a single source and cannot be cross-verified.
The more substantive uncertainty lies in continuity: whether NVIDIA plans to continuously open-source and maintain all of Run:ai's code, or only some components, is not made clear by the source. This directly bears on whether the "open foundation" is a long-term commitment or a one-time release.
Also unresolved: the implementation timelines and complete functional scope of each ecosystem partner's integration, the specific magnitude of GPU utilization and throughput improvements after open-sourcing, the foundation affiliation of the three components other than KAI Scheduler, and the migration and pricing arrangements for existing paid customers.
This article is based on a release note from the NVIDIA Developer Blog (October 13, 2026). The open-source release, the four modules, KAI Scheduler's Apache 2.0 license and CNCF Sandbox affiliation, the list of eight partners, the code repository and the NGC Community Product's availability and support boundaries are all handled as stated by the source; the specific approaches of partner integrations, the continuation of the commercial product line, and the throughput and utilization benefits brought by open-sourcing are source claims and have not been independently verified; the judgment regarding NVIDIA's ecosystem strategic intent is an inference. The release date comes from a single source and cannot be cross-verified.
What is truly noteworthy about this release is not "free" but that the control point of the orchestration layer has been moved to the community repository: all development of KAI Scheduler takes place in the community, while enterprise support is left on the commercial product line. For adopters, the question has shifted from "whether to buy" to "use the open-source version or buy support," and this choice currently still lacks information on migration paths and quantified benefits.
Primary report
Primary source