Intel AI Platform Optimized to Support Microsoft’s Phi-3 AI Model on the Day of Release

Intel and Microsoft collaborate to support multiple Phi-3 models across their data center platforms, AI PCs and edge solutions

Recently, Intel has verified and optimized its AI product portfolio across clients, edge, and data centers for Microsoft's multiple Phi-3 family open models. The small open models of the Phi-3 family can run on low-computing hardware and are easier to fine-tune to meet specific user requirements, allowing developers to easily build applications that run locally. The product portfolio that supports this model includes Intel® Xeon® processors and Intel® Gaudi AI accelerators for data center applications, as well as Intel® Core™ Ultra processors and Intel RTX™ graphics cards for clients.

Intel® Core™ Ultra processors support Phi-3 family open models

“We provide customers and developers with powerful AI solutions that use the latest AI models and software in the industry. Deep collaboration between Intel and AI software ecosystem leaders like Microsoft is key to making AI ubiquitous. We are very excited to work with Microsoft to enable a range of Intel hardware products covering data centers, edge and clients to support multiple new Phi-3 models.”

—Pallavi Mahajan, Vice President and General Manager of Data Center and AI Software at Intel

Why it matters: To drive AI everywhere, Intel continues to build its AI software ecosystem by collaborating with AI leaders and innovators.

Intel and Microsoft collaborated to support the Phi-3 model on the day of its release with their product portfolio, including Intel central processing units (CPUs), graphics processing units (GPUs), and Intel Gaudi accelerators. In addition, Intel and Microsoft jointly designed an easy-to-use deep learning optimization software suite in DeepSpeed ​​- the accelerator abstraction layer, and expanded automatic tensor parallel support for Phi-3 and other models on Hugging Face.

The size of the Phi-3 model is ideal for on-device reasoning, and enables lightweight model development such as fine-tuning or customization on AI PCs and edge devices. At the same time, Intel client hardware is accelerated through a comprehensive software framework and tools, including PyTorch and the Intel® PyTorch Extension Pack for local R&D, and the OpenVINO™ toolkit for model development and reasoning.

In the future, Intel will continue to meet the generative AI needs of enterprise users and will continue to support and optimize software for Phi-3 and other industry-leading language models.