Dell Technologies and Meta are committed to fostering a broad, open and inclusive AI community for developers and technologists. Open-source technologies are transforming the AI industry, and the collaboration between Dell and Meta continues to deliver new possibilities for enterprise AI.
As an industry leader, Dell is proud to be the first on-premises infrastructure provider to offer standardized AI solutions optimized for the Llama ecosystem.¹ Together with Meta, we are delivering new Dell AI Solutions with the latest Llama 4 models and Llama Stack distribution with the Dell AI Factory. This enables organizations to develop advanced solutions without the limitations of public cloud-only environments.
Pursuing New Opportunities with Agentic AI
Agentic AI marks a major evolution in artificial intelligence. Unlike traditional AI with fixed rules or generative AI that creates content, agentic AI l and takes initiative to achieve specific outcomes. It unlocks smarter automation by streamlining end-to-end processes, enhances decision-making with faster and more accurate operations, and drives new growth.
Agentic AI represents a promising shift towards more intelligent and autonomous However, it requires careful planning and robust frameworks to deliver meaningful results. Building trust through transparent, explainable AI frameworks is essential to ensure accountability. Tight integration with existing systems is also required to minimize workflow disruptions, and effective data management is necessary for agents to function optimally.
This is where Llama Stack comes in. Llama Stack is a comprehensive open-source platform that provides a uniform set of tools and APIs for building, scaling and deploying AI applications. It enables developers to create, integrate and orchestrate multiple AI services and capabilities into composable multi-step workflows that deliver repeatable outcomes.
Dell’s validated reference architecture pairs Llama Stack software building blocks with Dell PowerEdge servers to deliver consistent performance and scalability. Developers can confidently use Llama Stack tools and Dell best practices for model training, fine-tuning and inferencing in a production deployment, maintaining efficiency every step of the way. This greatly simplifies application development with the latest Llama 4 models.
Key benefits of using Llama Stack with Dell AI Factory include:
-
- Support demanding workloads with a scalable, adaptable architecture for organizations of all sizes.
- Simplified deployment for faster AI integration with a of AI toolchain, inferencing engine and vector database configurations, accelerating time-to-value.
- Protect AI models, data sources and outputs with Dell PowerProtect data protection, which ensures that AI data and applications are recoverable in the event of data loss, corruption, operational issues, or the growing threat of cyberattacks.
- Support diverse use cases including enterprise search with multimodal retrieval-augmented generation (RAG), lengthy document analysis, complex process automation with AI agents, enhanced code generation and AI-powered language translation.
- Cost-efficiency with right-sized Dell infrastructure and Meta’s models to reduce expenses while maximizing ROI.
Taking Advantage of Llama 4 Innovation
The latest Llama 4 models are designed to improve efficiency and deliver faster, more powerful AI processing. Built with an advanced mixture-of-experts (MoE) architecture, the models can leverage the strengths of multiple “experts”, which are smaller, specialized neural networks within the larger MoE model. Each expert is trained to be particularly good at handling certain types of data or tasks. The models activate only a subset of experts per input, reducing computational costs while maintaining high performance.
There are two new models: Llama 4 Scout (17B active parameters) and Llama 4 Maverick (17B active parameters, 128 experts). The former can fit on a single NVIDIA H100 GPU while the latter fits on a single server with multiple H100 GPUs. Engineered for native multimodality, these models use to integrate text and vision tokens, enabling joint pre-training on large-scale data and featuring an improved vision encoder for enhanced performance. In addition, Meta released Llama Guard 4, a multimodal safety classifier with 12B parameters capable of analyzing text and multiple images to classify content for safety in both prompts and responses.
Highlights of Llama 4 include:
-
- Multimodal capabilities that combine text, image and video processing, opening doors for powerful AI applications across industries.
- Extended context lengths to analyze large datasets like legal or scientific documents with greater accuracy.
- Optimized computational efficiency by using an advanced mixture-of-experts (MoE) architecture that activates only a few experts per input.
- Customizable models that can be fine-tuned to enhance application-specific performance.
These advancements position Llama 4 as an essential tool for enterprises looking to build secure, scalable AI frameworks for agent-based applications. A variety of Dell AI Factory infrastructure configurations are supported using PowerEdge R-Series and XE-Series servers with advanced GPUs from NVIDIA and AMD. PowerEdge, with its high-speed in-node and cross-node network fabric, is an ideal platform for both single-node and multi-node LLM inferencing with Llama models.
The following configurations are validated with Llama 4 on the Dell Enterprise Hub:
Get Started Today
Ready to take your AI strategy to the next level? Dell makes it easy to explore and adopt its AI solutions with Llama. Developers can download ready-to-deploy containers with Llama 4 models directly from Dell Enterprise Hub on Hugging Face for seamless on-premises deployment. In addition, Dell’s AI services team is standing by to provide expert assistance based on our extensive, global experience deploying AI at scale. Contact your Dell sales rep to setup a free half-day facilitated workshop to determine how your business can benefit from AI.
For those attending Dell Technologies World, stop by the Dell Technologies booth (#833) to see the Llama Stack demo and talk with experts about our joint solutions.
1 Based on internal analysis, September 2024.
