I spent most of the last decade in the Midwest (hello, Chicagoland!) but now live in the Seattle area and work at the Microsoft campus in Redmond. Before moving here, I’d visited the Redmond campus dozens of times as a Microsoft partner and customer, but had only basic airport-campus-hotel-restaurant navigation skills. Relocating a family of four meant a lot of exploring to find my way to grocery stores (what do you mean there’s no Mariano’s in Seattle?), schools, doctors, dentists, etc. I couldn’t have done it without GPS: Just plug in the destination and let it guide us, turn by turn, on the quickest, safest route, avoiding traffic and roadwork along the way.
The hallmark of good tech is taking something complex and stressful (ever been lost in an unfamiliar city?) and making it simple. Looking over the list of new AI capabilities we announced in just the past month, it’s easy to see why turn-by-turn guidance is needed in this rapidly expanding AI landscape. Azure AI Foundry is like GPS for your AI ambitions: it simplifies navigating the insane pace of innovation in the AI landscape so you see where you are today, evaluate what’s new, and plan, execute, and manage your growing AI portfolio while optimizing costs. All with the confidence that it always has the latest AI thing.
This wayfinding provides dexterity to help your teams stay nimble as AI presents new strategic opportunities and efficiencies. Here are just a few new capabilities in Azure AI Foundry this month that show the value of managing your AI portfolio on a single do-it-all platform.
Models: More is more
As I write this there are 1,917 models in the Azure AI Foundry model catalog. Choosing one—or three, since many AI apps now use multiple models—is easier than you might think when facing an abundance of choice. On Azure AI Foundry, your team can quickly evaluate models on quality, cost, and throughput, or filter by industry, inference tasks like translation or text generation, or other key features to quickly get a short list of models to consider. New tools like the Evaluation API in Azure OpenAI Service give developers the ability to run automated comparisons between OpenAI models, quantify quality improvements, and accelerate the iteration cycle. Together with the Stored Completions API, the Evaluation API provides a programmatic solution for refining and evaluating AI models with accuracy and control.
That’s important when you consider we announced eight new Azure OpenAI Service models in one week—the new GPT-4.1 model series, three advanced audio models, and the o3 and o4-mini reasoning models. You can fine-tune GPT-4.1 and GPT-4.1-mini to adapt them to your unique business requirements. And last week we announced GPT-image-1, the latest and most advanced image generation model, as well as the latest GPT-4o-realtime-preview and GPT-4o-mini-realtime-preview API updates including WebRTC support, semantic voice activity detection and advanced input noise reduction and more.
Check out how Costar Group uses Azure OpenAI Service models to transform their customers’ experience of finding the right home.
That much newness in one week might seem like an anxiety accelerator—especially if your team just built an app using an earlier model—but Azure AI Foundry was designed for the progressive, iterative flow of AI development. It’s easy for teams to quickly evaluate new model pros and cons, drop the model in the playground for some hands-on testing, determine if it’s better, faster, cheaper, etc., and deploy it if desired. Simple. And we intend to enable OpenAI models in Azure AI Foundry the same day as OpenAI announces them. They are backward-compatible and can be deployed with a single click or by adding a new model with a few simple lines of code.
Azure OpenAI accounts for 35 of the 1,917 models in our catalog, but is not alone in introducing new models. Also new in our model catalog this month:
The first models in the Meta Llama 4 herd arrived, including Llama 4 Scout, which specializes in condensing or analyzing extensive information, like analyzing all documents in an enterprise SharePoint library to answer a specific query, or reading a multi-thousand-page technical manual to provide troubleshooting advice. And Llama 4 Maverick, which specializes in image and text understanding with support for 12 languages for precise image understanding and creative writing.
An incredible amount of knowledge sits in file cabinets around the world, invisible to AI. Mistral OCR, a state-of-the-art Optical Character Recognition (OCR) model, can read and understand the structure and nuance of hard copy documents. It can extract equations, images, tables, and multilingual text from scanned contracts, handwritten forms, research papers, regulatory filings, and more at scale, making all that knowledge accessible for AI to reason over and your team to learn from and act on.
Optimized for the enterprise, Cohere’s Command A and Embed 4 models form a powerhouse for advanced apps like extracting insights from thousands of financial filings, earnings call transcripts, and market research.
DeepSeek got the world’s attention with the release of its R1 model earlier this year. It’s a powerful and capable, low-cost model with some safety challenges requiring careful management. The Microsoft AI team developed its own DeepSeek R1 variant, MAI-DS-R1, that outperforms DeepSeek R1 on responses to blocked topics and reducing harmful content while maintaining the same reasoning capabilities. You can find MAI-DS-R1 in Foundry today.
Go to market faster with Azure AI Services
Azure AI Services are efficient time-to-market accelerators that help your developers build intelligent apps without needing direct AI or data science skills. Consider agent building. Until recently, creating agents was complicated and consumed a lot of developer time. With Azure AI Agent Service in Azure AI Foundry, what originally took hundreds of lines of code can now be done in just a few lines of code. Azure AI Agent Service is used by tens of thousands of organizations to create and deploy agents. Customers like Fujitsu, who built an agent to help prioritize sales leads, and NTT DATA, which built an agent that speeds proposal creation and surfaces client insights.
Azure AI Search is another great example. This revolutionary retrieval system provides highly relevant search results at scale. It connects corporate knowledge into AI applications to power experiences like UBS’s Legal AI Assistant (LAIA). This assistant helps their legal team perform research across multiple languages and 26 million documents to quickly find related clauses to a case or provisions across investment funds.
We recently announced several improvements enabling your team to upgrade to the newest retrieval and storage features without disrupting service. These new features make Azure AI Search one of the most cost-effective options on the market, delivering an enterprise-ready, full-featured retrieval system with advanced search technology without sacrificing cost or performance. The result is empowering your app to deliver high-quality experiences for every user and interaction with no compromises.
Coca?Cola turned its iconic Santa into a globe?trotting conversationalist in just 60 days thanks to an AI?ready foundation on Azure. Using Azure AI Speech in Azure AI Foundry, the team built a real?time model that let Santa chat with more than a million people in 26 languages, powering an immersive Create Real Magic holiday campaign across 43 markets. It’s a masterclass in how Azure unlocks lightning?fast innovation at global scale.
Get there with Azure AI Foundry
Organizations built new muscle around resiliency and accelerating technology adoption during the pandemic and that conditioning is paying off as the AI era shifts the world again, reshaping what is possible. Efficiently navigating this fast-changing landscape is essential in staying competitive and building differentiation in a busy marketplace. Our intent is for Azure AI Foundry to be your trusted guide, simplifying complex processes and empowering your team to achieve more. With the latest advancements and a robust platform, you’re equipped to harness the full potential of AI, driving innovation and efficiency across your organization. Stay ahead of the curve with Azure AI Foundry—your GPS for the future of AI.
About Jessica
Jessica leads Azure Data, AI, and Digital Applications at Microsoft. Find Jessica’s blog posts here and be sure to follow Jessica on LinkedIn.