Upbound Submits Modelplane to CNCF Sandbox
The open source AI inference control plane seeks a neutral community home for managing models and GPU infrastructure across clusters.
Updated on Nov. 10, 2026 in Artificial Intelligence
Upbound, the company behind Crossplane, submitted Modelplane, an open source control plane for AI inference, to the Cloud Native Computing Foundation Sandbox on November 10, 2026. The project was demonstrated running inference across AWS and Vultr infrastructure at KubeCon + CloudNativeCon North America.
Why it matters
Teams are increasingly running different models, serving technologies, and GPU infrastructure across multiple clusters and providers, creating a need for coordination across that fleet. Modelplane is positioned to address that coordination problem under a neutral community governance model if accepted.
Modelplane integrates with NVIDIA Dynamo, using Grove and the KAI Scheduler to coordinate multi-node workloads and ModelExpress to transfer model weights directly between GPUs. The project is available under the Apache 2.0 license, and the stub does not report performance benchmarks for these integrations.
The players
Upbound
Upbound is the San Mateo, California-based company behind the open source Crossplane infrastructure project and now Modelplane.
Cloud Native Computing Foundation
The Cloud Native Computing Foundation is a Linux Foundation-hosted open source organization that stewards Kubernetes and related cloud-native projects, and its Sandbox is an entry stage for early-stage projects.
NVIDIA
NVIDIA is a semiconductor company whose GPUs and Dynamo inference framework Modelplane integrates with for multi-node AI serving.
The details
Platform and AI teams specify which models to run and what infrastructure those models need, and Modelplane coordinates deployments, serving engines, and GPU capacity across a fleet of Kubernetes clusters and cloud providers. Upbound uses Modelplane internally to run its own chat, coding assistant, and security agent workloads.
Timeline
November 10, 2026: Upbound submits Modelplane to the CNCF Sandbox.
Demonstration of inference across AWS and Vultr infrastructure at KubeCon + CloudNativeCon North America.
The Tech Race
Modelplane extends the control-plane model Upbound built with Crossplane, which has reached 100 million downloads, into the specialized domain of AI inference and GPU fleet management. The move positions Upbound against other infrastructure vendors competing to own the orchestration layer for production AI workloads.
Engineering teams running models across multiple Kubernetes clusters and cloud providers could use a single control plane to coordinate deployments and GPU capacity, if the project is accepted and adopted. Developers should note that the project is still in the submission stage and is not yet a CNCF-accepted project.
The takeaway
Managing AI models and GPU capacity across multiple clusters and clouds is becoming a platform-level problem, and teams evaluating inference infrastructure should watch how neutral governance efforts like Modelplane's Sandbox submission develop.
Further reading
Explore more Artificial Intelligence coverage for context on how inference infrastructure is evolving.
More information
Review the source code and documentation in the Modelplane open source project repository.
Source note: This article includes information reported by The Manila times.
