Скачать или смотреть From Containers to Models: Docker’s AI Evolution with Model Runner and Model format by Kevin Wittek

The current AI wave has taken the industry by storm—and often, it’s easier to ride it than to stand against it. While container technology, especially Docker, has long played a role in AI development, it often felt like an awkward fit. The pieces were there, but assembling them required significant effort.
Building on the work of the open source community, we’ve now integrated powerful inference engines like llama.cpp directly into Docker tooling, including Docker Desktop and Docker CE. To further streamline AI workflows, we’re introducing the "Models as OCI Artifacts Specification"—a new subset of the OCI Artifact format, purpose-built for distributing LLM artifacts. And of course everything done as OSS ourselves.
In this session, you’ll learn how to leverage these new Docker capabilities for GenAI development. I'll show you how to package, publish, and consume your own custom models in the Docker Model format, making them easily accessible via any container registry that supports OCI Artifacts.