NVIDIA Open Sources Run:ai Platform with Four Modular Components

NVIDIA released the full Run:ai platform as open source on October 13, 2026, as part of the NVIDIA DSX AI Factory Platform. The release includes four modular components that can be adopted independently: KAI Scheduler, GPU Fractioning Operator, KAI Resource Management, and Workload Map. Source code is hosted in the NVIDIA DSX AI Factory GitHub organization, and a Helm chart for the integrated stack is available on NVIDIA NGC as a Community Product. No paid Run:ai subscription or NVIDIA support is required for the Community Product, though downloading from NGC requires acceptance of the NVIDIA Software License Agreement and Product-Specific Terms.

Components and Capabilities

KAI Scheduler is licensed under Apache-2.0 and has been contributed to CNCF as a Sandbox project. It supports gang scheduling, fair-share, topology awareness, GPU sharing, and dynamic resource allocation. The GPU Fractioning Operator uses CUDA 13.4 native fractioning and provides a metrics exporter for per-pod fraction visibility. It works with any scheduler but integrates tightly with KAI Scheduler. KAI Resource Management adds org-level projects, hierarchy, quotas, and node-pool grouping, enabling quota guarantees and sharing across projects and departments. Workload Map defines a consistent workload-management contract, providing a standardized contract for workload lifecycle management.

Ecosystem Integration

Eight DSX ecosystem partners are integrating the open source platform. Crusoe combines Run:ai with its Managed Kubernetes offering. Mirantis extends k0rdent AI with Run:ai for a metal-to-model AI factory stack. Nebius offers Run:ai as a one-click deployment in Nebius AI Cloud. OpenNebula integrates Run:ai via OneKS for sovereign AI clouds. Rafay provides Run:ai as a fully managed service with automated stack provisioning. Saturn Cloud uses Run:ai for white-label inference services with token-based metering. Spectro Cloud embeds Run:ai in PaletteAI for lifecycle-automated deployments. vCluster supplies pre-validated templates that deploy Run:ai per tenant.

Community and Commercial Paths

Developers can contribute to the code via the respective GitHub repositories, and community support is available through the CNCF Slack channel #kai-scheduler and GitHub issues. Enterprise AI Factory customers can still obtain Run:ai via commercial NVIDIA offerings. This open source release gives inference engineers and edge AI developers direct access to scheduling and GPU fractioning tools that were previously part of a proprietary platform, with the flexibility to adopt components individually or as a full stack.