News

Microsoft, Meta Find Common Ground in Llama LLM

• By David Ramel
• 07/26/2023

Microsoft, a consistent runner-up in the cloud computing contest, has recently forged closer connections with Meta, formerly known as Facebook, particularly in the domain of generative AI.

Redmond appears to be strengthening its collaboration with Meta to gain an advantage in the pivotal field of AI, a technology paramount in the contemporary IT landscape.

The AI partnership was unveiled last week during Microsoft's Inspire event, encompassing support for Meta's Llama 2 line of large language models (LLMs) on Azure and Windows. Llama 2, which is currently accessible online for experimentation, distinguishes itself as an open-source LLM, in contrast to proprietary models like GPT-4 from Microsoft's partner OpenAI, which are being integrated into various commercial products and services.

That support includes Microsoft being a preferred partner with Meta as it rolls out commercial wares based on the new LLM version.

"This announcement means that Azure customers will be able to easily fine-tune and deploy the 7B-parameter, 13B-parameter, and 70B-parameter Llama 2 models easily and safely on Azure," Microsoft said last week. "In addition, Llama 2 will be optimized to run locally on Windows -- enabling Windows developers to take advantage of Llama 2 by targeting the Direct ML execution provider through the ONNX runtime."

In its own announcement, Meta said:

• Llama 2 is free for research and commercial use.
• We're opening access to Llama 2 with the support of a broad set of companies and people across tech, academia, and policy who also believe in an open innovation approach to today's AI technologies.
• We're committed to building responsibly and are providing resources to help those who use Llama 2 do so too.

In another post, Microsoft noted Llama 2 is the latest addition to the Azure AI model catalog, a preview project that aims to serves as a hub of foundation models to help developers and machine learning (ML) professionals more easily discover, evaluate, customize, and deploy pre-built large AI models at scale.

Windows devs, meanwhile, must go to the Llama 2 ONNX GitHub Request Form.