Azure AI Foundry is like GPS in your AI ambitions: it simplifies navigating the insane tempo of innovation within the AI panorama so that you see the place you’re as we speak, consider what’s new, and plan, execute, and handle your rising AI portfolio.
I spent a lot of the final decade within the Midwest (whats up, Chicagoland!) however now dwell within the Seattle space and work on the Microsoft campus in Redmond. Earlier than shifting right here, I’d visited the Redmond campus dozens of instances as a Microsoft accomplice and buyer, however had solely primary airport-campus-hotel-restaurant navigation expertise. Relocating a household of 4 meant a variety of exploring to search out my technique to grocery shops (what do you imply there’s no Mariano’s in Seattle?), faculties, docs, dentists, and so forth. I couldn’t have executed it with out GPS: Simply plug within the vacation spot and let it information us, flip by flip, on the quickest, most secure route, avoiding site visitors and roadwork alongside the best way.
The hallmark of excellent tech is taking one thing advanced and tense (ever been misplaced in an unfamiliar metropolis?) and making it easy. Trying over the record of recent AI capabilities we introduced in simply the previous month, it’s straightforward to see why turn-by-turn steering is required on this quickly increasing AI panorama. Azure AI Foundry is like GPS in your AI ambitions: it simplifies navigating the insane tempo of innovation within the AI panorama so that you see the place you’re as we speak, consider what’s new, and plan, execute, and handle your rising AI portfolio whereas optimizing prices. All with the arrogance that it all the time has the most recent AI factor.
This wayfinding offers dexterity to assist your groups keep nimble as AI presents new strategic alternatives and efficiencies. Listed here are just some new capabilities in Azure AI Foundry this month that present the worth of managing your AI portfolio on a single do-it-all platform.
Fashions: Extra is extra
As I write this there are 1,917 fashions within the Azure AI Foundry mannequin catalog. Selecting one—or three, since many AI apps now use a number of fashions—is simpler than you may assume when dealing with an abundance of alternative. On Azure AI Foundry, your crew can shortly consider fashions on high quality, price, and throughput, or filter by business, inference duties like translation or textual content era, or different key options to shortly get a brief record of fashions to contemplate. New instruments just like the Analysis API in Azure OpenAI Service give builders the power to run automated comparisons between OpenAI fashions, quantify high quality enhancements, and speed up the iteration cycle. Along with the Saved Completions API, the Analysis API offers a programmatic answer for refining and evaluating AI fashions with accuracy and management.
That’s necessary when you think about we introduced eight new Azure OpenAI Service fashions in a single week—the brand new GPT-4.1 mannequin sequence, three superior audio fashions, and the o3 and o4-mini reasoning fashions. You may fine-tune GPT-4.1 and GPT-4.1-mini to adapt them to your distinctive enterprise necessities. And final week we introduced GPT-image-1, the most recent and most superior picture era mannequin, in addition to the most recent GPT-4o-realtime-preview and GPT-4o-mini-realtime-preview API updates together with WebRTC assist, semantic voice exercise detection and superior enter noise discount and extra.
Take a look at how Costar Group makes use of Azure OpenAI Service fashions to rework their prospects’ expertise of discovering the fitting dwelling.
That a lot newness in a single week may look like an nervousness accelerator—particularly in case your crew simply constructed an app utilizing an earlier mannequin—however Azure AI Foundry was designed for the progressive, iterative circulation of AI improvement. It’s straightforward for groups to shortly consider new mannequin execs and cons, drop the mannequin within the playground for some hands-on testing, decide if it’s higher, sooner, cheaper, and so forth., and deploy it if desired. Easy. And we intend to allow OpenAI fashions in Azure AI Foundry the identical day as OpenAI pronounces them. They’re backward-compatible and will be deployed with a single click on or by including a brand new mannequin with just a few easy traces of code.
Azure OpenAI accounts for 35 of the 1,917 fashions in our catalog, however shouldn’t be alone in introducing new fashions. Additionally new in our mannequin catalog this month:
The first fashions within the Meta Llama 4 herd arrived, together with Llama 4 Scout, which makes a speciality of condensing or analyzing intensive info, like analyzing all paperwork in an enterprise SharePoint library to reply a particular question, or studying a multi-thousand-page technical guide to supply troubleshooting recommendation. And Llama 4 Maverick, which specializes in picture and textual content understanding with assist for 12 languages for exact picture understanding and artistic writing.
An unbelievable quantity of information sits in file cupboards around the globe, invisible to AI. Mistral OCR, a state-of-the-art Optical Character Recognition (OCR) mannequin, can learn and perceive the construction and nuance of onerous copy paperwork. It could possibly extract equations, pictures, tables, and multilingual textual content from scanned contracts, handwritten varieties, analysis papers, regulatory filings, and extra at scale, making all that information accessible for AI to motive over and your crew to be taught from and act on.
Optimized for the enterprise, Cohere’s Command A and Embed 4 fashions kind a powerhouse for superior apps like extracting insights from 1000’s of monetary filings, earnings name transcripts, and market analysis.
Deep Search obtained the world’s consideration with the discharge of its R1 mannequin earlier this yr. It’s a robust and succesful, low-cost mannequin with some security challenges requiring cautious administration. The Microsoft AI crew developed its personal Deep Search R1 variant, MAI-DS-R1, that outperforms Deep Search R1 on responses to blocked subjects and lowering dangerous content material whereas sustaining the identical reasoning capabilities. You’ll find MAI-DS-R1 in Foundry as we speak.
Go to market sooner with Azure AI Providers
Azure AI Providers are environment friendly time-to-market accelerators that assist your builders construct clever apps without having direct AI or knowledge science expertise. Take into account agent constructing. Till just lately, creating brokers was sophisticated and consumed a variety of developer time. With Azure AI Agent Service in Azure AI Foundry, what initially took lots of of traces of code can now be executed in just some traces of code. Azure AI Agent Service is utilized by tens of 1000’s of organizations to create and deploy brokers. Prospects like Fujitsu, who constructed an agent to assist prioritize gross sales leads, and NTT DATA, which constructed an agent that speeds proposal creation and surfaces shopper insights.
Azure AI Search is one other nice instance. This revolutionary retrieval system offers extremely related search outcomes at scale. It connects company information into AI purposes to energy experiences like UBS’s Authorized AI Assistant (LAIA). This assistant helps their authorized crew carry out analysis throughout a number of languages and 26 million paperwork to shortly discover associated clauses to a case or provisions throughout funding funds.
We just lately introduced a number of enhancements enabling your crew to improve to the latest retrieval and storage options with out disrupting service. These new options make Azure AI Search probably the most cost-effective choices in the marketplace, delivering an enterprise-ready, full-featured retrieval system with superior search expertise with out sacrificing price or efficiency. The result’s empowering your app to ship high-quality experiences for each person and interplay with no compromises.
Coca‑Cola turned its iconic Santa right into a globe‑trotting conversationalist in simply 60 days due to an AI‑prepared basis on Azure. Utilizing Azure AI Speech in Azure AI Foundry, the crew constructed an actual‑time mannequin that permit Santa chat with greater than 1,000,000 individuals in 26 languages, powering an immersive Create Actual Magic vacation marketing campaign throughout 43 markets. It’s a masterclass in how Azure unlocks lightning‑quick innovation at international scale.
Get there with Azure AI Foundry
Organizations constructed new muscle round resiliency and accelerating expertise adoption throughout the pandemic and that conditioning is paying off because the AI period shifts the world once more, reshaping what is feasible. Effectively navigating this fast-changing panorama is crucial in staying aggressive and constructing differentiation in a busy market. Our intent is for Azure AI Foundry to be your trusted information, simplifying advanced processes and empowering your crew to realize extra. With the most recent developments and a sturdy platform, you’re outfitted to harness the complete potential of AI, driving innovation and effectivity throughout your group. Keep forward of the curve with Azure AI Foundry—your GPS for the way forward for AI.
About Jessica
Jessica leads Azure Knowledge, AI, and Digital Functions at Microsoft. Discover Jessica’s weblog posts right here and you’ll want to comply with Jessica on LinkedIn.
