We’re excited to share the primary fashions within the Llama 4 herd can be found at this time in Azure AI Foundry and Azure Databricks.
We’re excited to share the primary fashions within the Llama 4 herd can be found at this time in Azure AI Foundry and Azure Databricks, which permits folks to construct extra personalised multimodal experiences. These fashions from Meta are designed to seamlessly combine textual content and imaginative and prescient tokens right into a unified mannequin spine. This progressive strategy permits builders to leverage Llama 4 fashions in functions that demand huge quantities of unlabeled textual content, picture, and video information, setting a brand new precedent in AI growth.
In the present day, we’re bringing Meta’s Llama 4 Scout and Maverick fashions into Azure AI Foundry as managed compute choices:
- Llama 4 Scout Fashions
- Llama-4-Scout-17B-16E
- Llama-4-Scout-17B-16E-Instruct
- Llama 4 Maverick Fashions
- Llama 4-Maverick-17B-128E-Instruct-FP8
Azure AI Foundry is designed for multi-agent use circumstances, enabling seamless collaboration between totally different AI brokers. This opens up new frontiers in AI functions, from complicated problem-solving to dynamic activity administration. Think about a workforce of AI brokers working collectively to investigate huge datasets, generate artistic content material, and supply real-time insights throughout a number of domains. The chances are limitless.
To accommodate a variety of use circumstances and developer wants, Llama 4 fashions are available each smaller and bigger choices. These fashions combine mitigations at each layer of growth, from pre-training to post-training. Tunable system-level mitigations protect builders from adversarial customers, empowering them to create useful, protected, and adaptable experiences for his or her Llama-supported functions.
Llama 4 Scout fashions: energy and precision
We’re sharing the primary fashions within the Llama 4 herd, which can allow folks to construct extra personalised multimodal experiences. In response to Meta, Llama 4 Scout is likely one of the finest multimodal fashions in its class and is extra highly effective than Meta’s Llama 3 fashions, whereas becoming in a single H100 GPU. And Llama4 Scout will increase the supported context size from 128K in Llama 3 to an industry-leading 10 million tokens. This opens up a world of prospects, together with multi-document summarization, parsing intensive consumer exercise for personalised duties, and reasoning over huge codebases.
Focused use circumstances embody summarization, personalization, and reasoning. Due to its lengthy context and environment friendly measurement, Llama 4 Scout shines in duties that require condensing or analyzing intensive info. It might generate summaries or reviews from extraordinarily prolonged inputs, personalize its responses utilizing detailed user-specific information (with out forgetting earlier particulars), and carry out complicated reasoning throughout giant data units.
For instance, Scout might analyze all paperwork in an enterprise SharePoint library to reply a selected question or learn a multi-thousand-page technical guide to offer troubleshooting recommendation. It’s designed to be a diligent “scout” that traverses huge info and returns the highlights or solutions you want.
Llama 4 Maverick fashions: innovation at scale
As a general-purpose LLM, Llama 4 Maverick incorporates 17 billion energetic parameters, 128 specialists, and 400 billion whole parameters, providing prime quality at a lower cost in comparison with Llama 3.3 70B. Maverick excels in picture and textual content understanding with assist for 12 languages, enabling the creation of subtle AI functions that bridge language limitations. Maverick is right for exact picture understanding and artistic writing, making it well-suited for common assistant and chat use circumstances. For builders, it provides state-of-the-art intelligence with excessive velocity, optimized for finest response high quality and tone.
Focused use circumstances embody optimized chat eventualities that require high-quality responses. Meta fine-tuned Llama 4 Maverick to be a superb conversational agent. It’s the flagship chat mannequin of the Meta Llama 4 household—consider it because the multilingual, multimodal counterpart to a ChatGPT-like assistant.
It’s significantly well-suited for interactive functions:
- Buyer assist bots that want to grasp photos customers add.
- AI artistic companions that may focus on and generate content material in numerous languages.
- Inner enterprise assistants that may assist workers by answering questions and dealing with wealthy media enter.
With Maverick, enterprises can construct high-quality AI assistants that converse naturally (and politely) with a worldwide consumer base and leverage visible context when wanted.
Architectural improvements in Llama 4: Multimodal early-fusion and MoE
In response to Meta, two key improvements set Llama 4 aside: native multimodal assist with early fusion and a sparse Combination of Consultants (MoE) design for effectivity and scale.
- Early-fusion multimodal transformer: Llama 4 makes use of an early fusion strategy, treating textual content, photos, and video frames as a single sequence of tokens from the beginning. This allows the mannequin to grasp and generate numerous media collectively. It excels at duties involving a number of modalities, similar to analyzing paperwork with diagrams or answering questions on a video’s transcript and visuals. For enterprises, this enables AI assistants to course of full reviews (textual content + graphics + video snippets) and supply built-in summaries or solutions.
- Slicing-edge Combination of Consultants (MoE) structure: To attain good efficiency with out incurring prohibitive computing bills, Llama 4 makes use of a sparse Combination of Consultants (MoE) structure. Primarily, which means that the mannequin includes quite a few knowledgeable sub-models, known as “specialists,” with solely a small subset energetic for any given enter token. This design not solely enhances coaching effectivity but in addition improves inference scalability. Consequently, the mannequin can deal with extra queries concurrently by distributing the computational load throughout numerous specialists, enabling deployment in manufacturing environments with out necessitating giant single-instance GPUs. The MoE structure permits Llama 4 to increase its capability with out escalating prices, providing a big benefit for enterprise implementations.
Dedication to security and finest practices
Meta constructed Llama 4 with one of the best practices outlined of their Developer Use Information: AI Protections. This consists of integrating mitigations at every layer of mannequin growth from pre-training to post-training and tunable system-level mitigations that protect builders from adversarial assaults. And, by making these fashions out there in Azure AI Foundry, they arrive with confirmed security and safety guardrails builders come to count on from Azure.
We empower builders to create useful, protected, and adaptable experiences for his or her Llama-supported functions. Discover the Llama 4 fashions now within the Azure AI Foundry Mannequin Catalog and in Azure Databricks and begin constructing with the most recent in multimodal, MoE-powered AI—backed by Meta’s analysis and Azure’s platform power.
The provision of Meta Llama 4 on Azure AI Foundry and thru Azure Databricks provides clients unparalleled flexibility in selecting the platform that most closely fits their wants. This seamless integration permits customers to harness superior AI capabilities, enhancing their functions with highly effective, safe, and adaptable options. We’re excited to see what you construct subsequent.
