AI Foundry is a Microsoft platform that helps to customize, deploy, and manage AI models, including Cognitive Services. We are working on integrating it and making it ACP2 compliant.

While we have made significant progress, we are collaborating with Microsoft to address some open topics to ensure a seamless experience. We will keep you updated on the progress and ensure you are among the first to know once AI Foundry is integrated.

You have a Developer package, but haven`t heard of the LLMOps Template?

The LLMOps Template brings to the table an easy to follow setup, ability for local development, evaluation flows, model monitoring as well as tracing capabilities. It currently provides an easy-to-use setup RAG pipeline, as well as a framework for including multimodality.

DeepSeek is a new open source model that is now available, this model supports efficient inference with reduced computational requirements.

Certain versions of Deepseek are available as a service though the model catalog (pay as you go, streamlined deployment).  Additionally, DeepSeek and other open source models can be imported and self-hosted on a Nvidia GPU via Hugging Face integration. 

Be mindful of the risks associated with importing and hosting such models.