Modelplane
A self-hosted, Kubernetes-based control plane, built on Crossplane, that provisions and schedules AI inference across any GPU infrastructure and exposes each deployed model behind one OpenAI-compatible endpoint.
Modelplane is software a team installs and runs in its own environment to orchestrate AI models, the serving engines running them, and the infrastructure underneath, across cloud, neocloud, and on-premises hardware, from a single GPU to disaggregated multi-node deployments. Built on the Crossplane project, it continuously reconciles a fleet toward a declared state: provisioning inference clusters, scheduling deployments onto compatible clusters, scaling replicas, caching model weights, and routing traffic.
The intended split of responsibility is that platform teams provision clusters and publish available hardware classes, while developers declare a model with a manifest and receive a unified, OpenAI-compatible endpoint, without either side needing to understand the other's job in detail. It is an early-stage project under active development, with APIs still subject to change, and is Apache 2.0 licensed.