Models from Meta integrated: Microsoft upgrades Azure with Llama 4 models
Developers of AI tools can use the new Llama 4 language models from Meta in Microsoft Azure. Azure AI Foundry is also getting new functions for AI agents.
(Image: VDB Photos/Shutterstock.com)
Microsoft has integrated the AI models Llama 4 Scout and Maverick into Azure AI Foundry and Azure Databricks. By providing the different variants of the language model, the US software company aims to meet the needs of developers of AI applications. According to the company, they have configurable protection mechanisms that developers can use to protect themselves against the misuse of their AI tools. This includes the filtering of training data, for example.
Llama 4 Scout writes reports and summaries
As Meta announced at the presentation of the Llama 4 family, Scout and Maverick can process both text and images. There are different experts within a model, but only those experts answer a query that the model considers useful. Microsoft cites the aggregation and analysis of information as well as the creation of reports and conclusions from a large amount of data as possible applications. In its blog, Microsoft cites the creation of a technical manual based on SharePoint documents as a concrete example.
Llama 4 Maverick, on the other hand, is suitable for developing interactive applications, such as chatbots for customer support, which are designed to handle images uploaded by users. The Maverick model can also be used to create internal company assistants and for the discussion and generation of creative content. According to Meta, Maverick has over 400 billion parameters, a context length of one million parameters and supports twelve different languages, including German, English, French, Italian and Spanish. The model can be run on an H100 host.
Azure AI Foundry gets new functions for AI agents
Microsoft also announced new functions in Azure AI Foundry for the development of AI agents in its blog. This includes an agent framework that extends the open-source development kit Semantic Kernel. It is designed to facilitate coordination between agents and require less code from developers. The Red Teaming Agent, which tests AI models for security risks and summarizes its assessment in a report, is available as a public preview. Microsoft has also released a preview of an extension for Visual Studio Code that allows agent-based AI tools to be tested and executed within the development environment.
(sfe)
