Microsoft gives Azure AI users a place to experiment with latest technologies

With the pace of innovation in AI accelerating, it’s important for developers to get early access to new tools and technologies. To that end, Microsoft has introduced a module inside its Azure AI Foundry service that aims to bring the latest innovations and research in AI to developers.

“We’ve witnessed a shift from unveiling a new model every 4–6 months to releasing breakthroughs every 4–6 days,” the company wrote in a blog post, adding that the rebranded module could act as a bridge between research and application building.

Called Labs, the rebranded module will enable developers to experiment with new models and explore the latest frameworks so that they can create new prototypes and collaborate with researchers and engineering teams by sharing feedback, the company wrote.

The rebranding exercise is a play to become more competitive in AI and cloud markets, said Charlie Dai, vice president and principal analyst at Forrester.

“By providing an integrated platform with advanced AI capabilities in the cloud ecosystem, Labs can attract developers and enterprises and drive adoption and differentiation, aiming to position Microsoft Azure as a leader in AI development,” Dai explained.

As part of Labs, Microsoft has added new large language models (LLMs) and agentic frameworks, such as Aurora, ExACT, Magentic-One, MatterSim, OmniParser V2, and TamGen.

While Aurora is a large-scale atmospheric model that provides high-resolution weather forecasts and air pollution predictions, MatterSim is a deep learning model for atomistic simulations and predicting material properties with high precision.

TamGen and OmniParser V2, on the other hand, can be used for drug designing and converting UI screenshots into structured elements to aid agents’ action generation.

Other resources, such as ExACT and Magentic-One are related to agents and agentic frameworks.

While ExACT is an open source project enabling agents to learn from past interactions to improve search efficiency, Magentic-One is a multi-agent framework, built atop its proprietary AutoGen framework, that aims to solve complex problems without human intervention.

Magentic-One has a multi-agent architecture in which one agent, the Orchestrator, directs four other agents to solve a given task.

Another addition to Labs is Microsoft’s recently released World and Human Action Model (WHAM) named Muse, which can understand gameplay and take actions inside a game.

These LLMs and agentic frameworks inside Labs will equip developers with AI tools that span the modelOps lifecycle, such as foundation models, support for prompt engineering, RAG, and governance capabilities, said Dai.

However, Microsoft is not alone and rival hyperscalers, such as Google and AWS, have also introduced similar services.

Google also has Google Labs that provide developers access to latest AI innovations such as Career Dreamer, Project Astra, Project Mariner, and NotebookLM among others.

While Project Astra is a research prototype exploring future capabilities of a universal AI assistant, Mariner is a research prototype exploring the future of human-agent interaction, starting with your browser.

AWS, on the other hand, also offers PartyRock — a mostly free low code tool for building generative AI applications with a new app search function and the ability to build document processing into applications.  

Last year, Microsoft had packaged its Azure AI Studio and other updates into a new service — Azure AI Foundry in response to enterprises’ need to develop, run, and manage generative AI applications.

Azure AI Foundry competes with rival offerings such as Google Vertex AI and AWS’ Amazon Bedrock. Chinese cloud providers, such as Alibaba, Tencent and Huawei also offer similar products in the form of PAI, TI, and ModelArts respectively, Dai said.