Meta and Microsoft join forces to bring Llama 2 language models to Azure and Windows

Devesh Beri

Looking for more info on AI, Bing Chat, Chat GPT, or Microsoft's Copilots? Check out our AI / Copilot page for the latest builds from all the channels, information on the program, links, and more!

Meta and Microsoft have announced their collaboration at Microsoft Inspire to introduce support for the Llama 2 family of large language models (LLMs) on Azure and Windows. Llama 2, developed by Meta, empowers developers and organizations to build AI-powered tools and experiences. This partnership reinforces both companies’ commitment to democratizing AI. Here’s Meta’s official statement.

Last week Meta announced that they are set to launch a commercial AI model and aim to compete with OpenAI and Google in the generative AI race.

Azure customers can now fine-tune and deploy the 7B, 13B, and 70B-parameter Llama 2 models on Azure, renowned for supporting open and frontier models. Additionally, Llama will be optimized to run locally on Windows, enabling Windows developers to integrate Llama into their applications seamlessly.

Microsoft’s partnership with Meta

Meta and Microsoft have a longstanding partnership in AI, including integrating ONNX Runtime with PyTorch. The latest collaboration expands Microsoft’s open model ecosystem and further solidifies Azure’s position as the world’s supercomputing platform for AI.

Azure’s purpose-built AI supercomputing platform, designed to handle demanding AI workloads, supports the Llama 2 models. Developers gain access to powerful tools for model training, fine-tuning, inference, and AI safety.

Integrating Llama 2 models in Windows elevates Windows as the go-to platform for building AI experiences tailored to specific needs, leveraging tools like Windows Subsystem for Linux, Windows Terminal, and Microsoft Visual Studio.

Expanding Azure AI model catalog and Windows availability

Llama 2 is the newest addition to Azure’s growing AI model catalog, enabling developers and ML professionals to quickly discover, evaluate, customize, and deploy pre-built large AI models at scale. The record eliminates infrastructure management hassles and supports optimization techniques like DeepSpeed and ONNX Runtime.

Windows developers can create new experiences using Llama 2 through the GitHub Repo. With Windows Subsystem for Linux and powerful GPUs, developers can fine-tune Llama 2 models on their Windows PCs.

Building responsibly with Azure

Microsoft’s responsible AI approach ensures AI safety through iterative, layered measures. Customers can test Llama 2 with their data, utilize prompt engineering and retrieval augmented generation (RAG) techniques, and leverage Azure AI Content Safety for a safer online experience.

Expanding the model catalog with Llama 2 and the partnership with Meta emphasize a responsible and open approach to AI.