Nvidia open-sources entire Run:ai AI and GPU management platform
The code is on GitHub and offered as a community product on NVIDIA NGC, without enterprise support, alongside four standalone components.
NVIDIA has released its entire Run:ai platform, which manages AI workloads and GPU resources, as open source. The release, dated October 19, 2026, covers the full platform plus four standalone core components, according to digitaltoday.co.kr.
Run:ai allocates GPU resources on demand in Kubernetes environments and manages AI work from data processing through training and inference. The standalone components include KAI Scheduler, which decides job placement so multiple teams share cluster resources fairly, and GPU Fractional Operator, which lets multiple jobs share a single GPU. KAI Resource Management sets and guarantees GPU quotas by department and project, while Workload Map defines various AI workload types running on Kubernetes in a consistent way.
The source code is available from the GitHub DSX AI Factory organization and as a community product on NVIDIA NGC. The NGC community product can be used without a paid subscription but does not include NVIDIA enterprise support.
NVIDIA also introduced two related open-source projects. Model Streamer loads model weights into GPU memory quickly to shorten inference startup time, and Grove manages multi-component inference workloads that must scale together through a single API.
Partners including Cruncho, Nebius, Mirantis, Rafay, SpectroCloud, SaturnCloud, OpenNebula Systems and vCluster have integrated Run:ai or offer it as a managed service. Nebius offers one-click deployment of Run:ai in its AI cloud.
Quick answers
Is Run:ai now open source?
Yes. NVIDIA released the entire Run:ai platform, along with four standalone core components, as open source.
Where can I get the Run:ai source code?
It is available from the GitHub DSX AI Factory organization and as a community product on NVIDIA NGC.
Does the NGC community edition include enterprise support?
No. It can be used without a paid subscription but does not include NVIDIA enterprise support.