Microsoft to Distribute Llama 2, Meta’s Open-Source Alternative to OpenAI LLM Models

Meta Llama 2

Microsoft announced at its Inspire conference yesterday that it’s expanding its AI partnership with Meta to accelerate innovation in that field. Microsoft will be Meta’s preferred partner to distribute Llama 2, the company’s next-gen and open-source large language model (LLM) designed for commercial use. 

Over the past couple of months, the AI conversation has mostly revolved around OpenAI and ChatGPT, the company’s partnership with Microsoft, and Google’s efforts to catch up. Microsoft used OpenAI technology to create its Bing Chatbot, Microsoft 365 Copilot, and Windows Copilot, and the company is also making OpenAI’s GPT-4 model available to commercial customers via its new Azure OpenAI service. 

With this expanded partnership with Meta, Microsoft shows that it won’t fully rely on OpenAI to offer AI tools to its customers. As for Meta, the company is also teaming up with Amazon and other providers to distribute its Llama 2 open-source AI model, even though Microsoft is currently the company’s “preferred” provider.

What is Meta’s Llama 2?

Llama 2 is the latest generation of Meta’s open-source large language model. It follows the public release of LLama 1 in February 2023, which according to the company received more than 100,000 requests from researchers to access it. 

Meta explained that it was trained on 40% more data than Llama 1 using publicly available online data sources, and it also has double the context length. If Llama 2 is available free of charge for research and commercial use, Meta does have an acceptable use policy to ensure that its AI model is used “fairly and responsibly.” 

The company believes that making Llama 2 open source can make its AI model more secure and trustworthy than the competition. “Our models outperform open-source chat models on most benchmarks we tested, and based on our human evaluations for helpfulness and safety, may be a suitable substitute for closed source models,” the company explained.

Meta offers different model sizes for its Llama 2 model
Meta offers different model sizes for its Llama 2 AI model (image credit: Meta)

Microsoft to support Llama 2 on Azure and Windows

Microsoft will make Llama 2 available on its Azure AI model catalog, which is currently in public preview. Azure customers will be able to leverage Llama 2 language models ranging from 7 billion to 70 billion parameters to build, train and deploy their AI workloads on Azure. 

In addition to making Meta’s LLama 2 AI model on Azure, Windows developers will also be able to use it locally via the official GitHub repository. “The inclusion of the Llama 2 models in Windows helps propel Windows as the best place for developers to build AI experiences tailored for their customers’ needs and unlock their ability to build using world-class tools like Windows Subsystem for Linux (WSL), Windows terminal, Microsoft Visual Studio and VS Code,” explained John Montgomery, CVP of Azure AI.

With this expanded AI partnership with Meta, Microsoft shows that it can play a unique role in the democratization of generative AI technologies by partnering with competitors and giving more choice to customers. “For years we’ve invested heavily in making Azure the place for responsible, cutting-edge AI innovation, whether customers are building their own models or using pre-built and customizable models from Microsoft, Meta, OpenAI and the open-source ecosystem,” Montgomery emphasized. 

Yesterday, Microsoft also revealed pricing details for its Microsoft 365 Copilot, which will use a $30 per person subscription model for commercial customers. The company also launched Bing Chat Enterprise in preview, which offers commercial data protection to customers worried about how ChatGPT and other AI chatbots can use their confidential data.