This week, Meta Platforms released Llama 2, the next generation of its open-source large language model that is free for research and commercial use. Llama 2’s pretrained and fine-tuned language models are available in sizes ranging from 7 to 70 billion parameters. Meta also named Microsoft Azure its “preferred partner for Llama 2,” offering it through the Azure AI model catalog for use with cloud-native tools that leverage content filtering and safety features. Meta says Llama 2 is “also optimized to run locally on Windows,” providing developers a seamless workflow across enterprise and consumer platforms.

While Meta couched its news release under the headline “Meta and Microsoft Introduce the Next Generation of Llama,” it noted Llama 2 is “also available through Amazon Web Services (AWS), Hugging Face, and other providers.”

Meta explained in a news announcement that it’s been “blown away by the huge demand for Llama 1 from researchers — with more than 100,000 requests for access to the large language model.”

Describing its relationship with Microsoft as “longstanding,” Meta says the two collaborated to advance use of the open source PyTorch framework for AI, and last month both joined the Partnership on AI’s framework for “Responsible Practices in Synthetic Media.” Microsoft’s pending acquisition of Activision Blizzard could hold compelling possibilities for the Meta Quest and the company’s VR ambitions.

“While it can’t match OpenAI’s GPT-4 in performance, Llama 2 apparently fares well for an open source model,” Ars Technica writes, quoting Nvidia Senior AI Scientist Jim Fan saying Llama “70B is close to GPT-3.5 on reasoning tasks, but there is a significant gap on coding benchmarks,” with the Meta LLM “on par or better than PaLM-540B … but still far behind GPT-4 and PaLM-2-L.”

MIT Technology Review writes that since OpenAI released “ChatGPT last November, tech companies have been racing to release models in hopes of overthrowing its supremacy,” with Meta “in the slow lane.”

When Microsoft and Google in February announced their AI chatbots, “Meta rolled out the first, smaller version of LLaMA, restricted to researchers, but it hopes that releasing LLaMA 2, and making it free for anyone to build commercial products on top of, will help it catch up.” (Note: Meta’s most current references stylize the spelling as Llama 2.)

Meta has released a research paper that details Llama 2’s technical specs.