Adoption
MatrixHub is an open-source, self-hosted, Hugging Face-compatible model registry for enterprise AI inference on Kubernetes.
Adopters
This list contains organizations that have agreed to be named publicly. If you would like your organization added or removed, open a pull request that updates this page.
DaoCloud runs MatrixHub on its Token Factory platform as a private model hub on Kubernetes, with Hugging Face-compatible access, intranet model caching, project-scoped repositories, and operational workflows.
External integration documentation
vLLM's MatrixHub documentation explains how to point HF_ENDPOINT at a MatrixHub instance so that vllm serve retrieves model weights over the internal network instead of the public Hugging Face Hub.
NVIDIA Dynamo's MatrixHub model-loading guide shows how to configure a Kubernetes deployment so that Dynamo frontend and worker components use MatrixHub to load pre-cached model weights.
llm-d's self-hosted registry guide explains how to configure HF_ENDPOINT in modelserver.env so that llm-d model-server Pods load pre-cached model weights from MatrixHub.
MatrixHub references
Add your organization
To add your organization, open a pull request that updates this page; contributions are welcome.