Meta✴ reported that its Llama AI large language models (LLMs) are approaching 350 million downloads, 10 times more than the same period last year. Roughly 20 million of those downloads were made in the past month alone, after the company released the Llama 3.1 language model, which Meta✴ says will allow it to compete directly with solutions from OpenAI and Anthropic.
Image source: Gerd Altmann / pixabay.com
Some of the largest cloud service providers partnering with Meta✴ saw monthly usage of Llama language models increase tenfold from January to July this year. It is also noted that from May to July, the use of Llama on the servers of its partners among cloud service providers more than doubled in terms of the number of tokens. In addition to Amazon Web Services (AWS) and Microsoft Azure, the company is partnering with Databricks, Dell, Google Cloud, Groq, Nvidia, IBM watsonx, Scale AI and Snowflake, among others, to make its LLMs more accessible to developers.
Meta✴ believes that the success of its language models is due to the fact that they are distributed under an open license. According to the company, the open distribution of its LLM has allowed it to “expand and diversify the AI ecosystem and provide developers with more choice.” When Meta✴ released Llama 3.1, CEO Mark Zuckerberg extolled the virtues of open-source AI, calling it “the way forward.” He also said the company is taking steps to make open-source AI an industry standard.
In its latest report, Meta✴ also revealed how its partners are using large language models. For example, AT&T uses Llama to refine user searches. One of the largest American food delivery companies, DoorDash, uses LLM to simplify the work of its software engineers. The language model is also used to generate live reactions and digital creatures in Niantic’s Peridot. Zoom, on the other hand, uses Llama, as well as other language models, to power an AI assistant that can summarize meetings and take smart notes.