NVIDIA opens Run:ai’s GPU orchestration platform
Run:ai’s platform code is now public, giving teams another way to share GPU capacity, while packaged downloads and support retain separate terms.
NVIDIA released the full Run:ai platform as open source on October 13, extending access beyond its previously opened scheduler.
The public repository includes control-plane services, cluster controllers and an installer; its top-level license is Apache 2.0. [1][2][3]
The practical problem is familiar: one team waits while another’s allocated GPUs sit idle. Run:ai can share spare capacity across team quotas and reclaim it when needed. Nebius says the platform is now available through its cloud application catalog. That is an adoption route, not an independent performance benchmark. [1][4]
Platform teams can evaluate the whole stack or individual components. Check the distribution as well as the code: NVIDIA’s NGC Community Product requires its software and product-specific terms, and excludes enterprise support. No paid Run:ai subscription is required for that community package. [1]
Take it further
Inspect the Run:ai source and installation guide For teams operating Kubernetes and GPU infrastructure. Deployments need compatible infrastructure and distribution-specific terms.
Sources & details 4 sources
Discuss in your voice app
Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.
Paste the brief into your voice app to discuss this reporting.Reporting sources
- NVIDIA · Platform release and distribution terms
October 13 original technical announcement. Full body read, including the distinction between source licenses, NGC Community Product terms and enterprise support.
- Run:ai · Public source repository
Actual public monorepo opened: control plane, Kubernetes controllers, shared libraries and installer. Availability inspected; no installation or operational validation performed.
- Run:ai · Repository license
Actual top-level LICENSE read: Apache License 2.0, NVIDIA copyright. This does not replace component-specific or packaged-distribution terms.
- Nebius · Its deployment offering
October 13 original adopter announcement by a commercial NVIDIA partner. It confirms its own catalog availability and describes quota borrowing. Prior commercial-version benchmarks are not presented here as new open-source release benchmarks.
Images & figures
Unused quota can be borrowed and reclaimed when its owner returns. Both arrays contain the same illustrative capacity; no throughput gain is measured here.
Euangelion · original explanatory figure · Source · Usage terms