Mirantis is adding new governance, routing, and metering capabilities to its k0rdent AI platform as enterprises and graphic processing unit (GPU) cloud operators look for more control over AI infrastructure moving from pilots into production.

The California-based company is expanding k0rdent AI beyond infrastructure management into the operational layer of AI delivery, where companies must track where models are running, how inference requests are routed, what those requests cost, and whether they comply with internal policies or sovereignty requirements.

The k0rdent AI Model Registry is built for storing and distributing large language models (LLMs), fine-tuned models, quantized builds, and other AI artifacts across distributed environments. Mirantis described it as an OCI-native registry designed specifically for AI model workflows, rather than traditional container images.

The k0rdent AI Inference Mesh is intended to route, meter, audit, and enforce policy on inference requests across models, regions, clusters, and providers. The goal is to give platform teams a clearer view of where AI traffic is going, how infrastructure is being consumed, and where compliance gaps may exist.

Mirantis also introduced k0rdent AI Inference Runtime, which the company said is designed to improve infrastructure efficiency by maximizing tokens per GPU-second.

“As organizations move AI projects from experimentation into production, infrastructure teams are increasingly confronting operational and governance challenges around model distribution, inference visibility, compliance enforcement, and GPU economics,” Kevin Kamel, VP of product development at Mirantis, noted in a statement.

The launch comes shortly after Iren agreed to acquire Mirantis for $625 million, with Mirantis expected to operate as a standalone subsidiary while supporting Iren’s AI cloud deployments. That deal positioned Mirantis’ bare metal GPU and Kubernetes stack as part of a broader push to bring AI infrastructure online faster.

Mirantis’ announcement also lands amid a rise of neocloud providers and distributed AI infrastructure. Neoclouds are building out high-performance networks for AI workloads, while companies such as Cast AI are trying to make GPU capacity more flexible across Kubernetes environments.