Open-source AI inference control plane Modelplane launches

Modelplane, an open-source project built on Kubernetes and Crossplane, has been launched. This tool provides a control plane that unifies GPU clusters distributed across cloud, Neocloud, and on-premise environments into a single Inference Fleet.

Modelplane operates on top of existing serving stacks such as vLLM, SGLang, and TensorRT-LLM, performing higher-level management functions beyond a single cluster.

Key Features:

• Model Placement and routing
• Autoscaling support
• Weight Caching management

This project is expected to be a useful tool for developers and companies looking to operate large-scale inference infrastructure on their own.