NVIDIA has open-sourced the Run:ai GPU orchestration platform under the Apache 2.0 license, featuring the KAI Scheduler, CUDA 13.4 GPU fractioning, and quota…
1 editorial report · 2 verified social mentions. The most authoritative report leads while later evidence completes the story.
Phancy Group and RiseUnion have launched ModelSphere, an open-source LLM deployment platform now available on GitHub under Apache 2.0. The project targets production serving with Kubernetes, Helm, SGLang, and vLLM, rather than a new model or chat app. Its README lists autoscaling, cache-aware request routing, a shared KV cache, support for several accelerator families, and recovery of in-flight generations after service interruptions. Those are project claims, not independent benchmark results. The quick start also needs a Kubernetes cluster, kubectl, Helm, helmfile, and helm-diff. The practical trade-off is clear: this is aimed at teams that want a control plane for model serving, but it adds another operational layer to run. Reddit searches found no current ModelSphere discussion, so pro
Open mention📰 NVIDIA Open Sources Run:ai Under Apache 2.0 for Kubernetes GPU Orchestration NVIDIA has released the full Run:ai GPU orchestration platform as open source under Apache 2.0. The release includes KAI Scheduler, CUDA 13.4 GPU fractioning, quota management, and a local Kind dev... 📰 Source: MarkTechPost 🔗 Link: https://www.marktechpost.com/2026/10/18/nvidia-runai-open-source-gpu-orchestration-kubernetes/ # AI # ArtificialIntelligence
Open mention