Meta launches first models of its latest Llama 3 LLM

Photo credit: Valeriya Zankovych / Shutterstock

Social media giant Meta has unveiled Meta Llama 3, the third version of its open-source large language model. The LLM is now available on a newly released version of the Meta AI virtual assistant, said founder and CEO Mark Zuckerberg in an Instagram video.

In addition, Llama 3 will soon be deployed on cloud computing platforms such as Amazon Web Services, Microsoft Azure, and Google Cloud, the company said in a statement. It will be supported by hardware platforms from the likes of Nvidia, Intel, and AMD.

Unlike Llama 2, which was released mid last year, Llama 3 uses pre-trained and refined language models with 8 billion and 70 billion parameters, respectively. Meta claims that the LLM has performed better on several benchmarks compared to Google’s Gemini Pro 1.5 and the Claude 3 Sonnet, among others.

Llama 3 is currently a text-based LLM supporting the English language, but Meta said that it aims to make it multilingual and multimodal “in the near future.” As Zuckerberg pointed out in his post, Meta AI is now integrated into the search boxes on top of WhatsApp, Instagram, Facebook, and Messenger, though the feature is not yet live in all countries.

Aside from OpenAI and Google, Meta’s heightened competition in the AI sphere also includes Chinese tech giants like Baidu, whose chatbot has breached 200 million users over a year after launch.

Meanwhile, startups like Reka AI have also launched language models that rival those from big players.

See also: Tether CEO wants to break Microsoft, AWS, Google’s hold on cloud computing