During Microsoft Inspire in July of this year, Microsoft and Meta jointly announced support for the Llama 2 family of large language models (LLMs). Llama 2 is a LLM that enables developers and organisations to create generative AI-powered tools. Both Meta and Microsoft are taking an open approach with Llama 2 in order to create a democratised AI future. Meta has chosen Microsoft as their preferred partner as they release their new Llama 2 version to commercial customers for the first time.

This new support for Llama 2 is the latest in the growing Azure AI model catalogue, which is currently in public preview. It serves as a central hub of foundation models and empowers machine learning experts to quickly and easily find, evaluate, customise, and deploy pre-built AI models. The catalogue means users need not manage all infrastructure dependencies when operationalising Llama 2. It offers turnkey support for fine-tuning models and evaluation, and significantly speeds up model fine-tuning.

AI safety

Responsible AI has been a major focus for Microsoft and the way they partner with other AI-focused organisations. They have invested deeply in Azure to make it the perfect platform for responsible AI innovation. Microsoft uses an interactive, layered approach to mitigate potentials risks brought forward by large language models. The Azure AI Content Safety service is one such tool that adds another level of protection to ensure safer online AI experiences. This partnership means Meta’s safety techniques will be combined with Azure AI Content Safety, leading to a layered safety approach to the deployments of Llama 2 models.

Working with Meta

Microsoft has worked with Meta as a close AI partner for a very long time, since the collaboration on integrating ONNX Runtime with Pytorch on Azure, and Azure being Meta’s choice as a strategic cloud provider. This announcement has brought them even closer to push for AI advancements and cements Microsoft’s open model ecosystem as the leading supercomputing platform for AI.

Azure’s purpose-built AI supercomputing platform was designed specifically to support the world’s leading AI organisations to build, train, and deploy advanced AI workloads. Llama 2 models becoming available with Azure AI means developers can take advantage of Azure AI’s powerful tooling for model training, tuning, and capabilities that support AI safely.