Meta is offering one of the best AI models for free. The company on Tuesday announced Llama 3.1, the most performant version of its large-scale language model yet, which will be free and open source, meaning anyone can easily access cutting-edge artificial intelligence models.
Meta CEO Mark Zuckerberg emphasized in an open letter on Tuesday that the company is taking a “different approach” by making its latest AI models “open source,” and is “actively building partnerships so more companies in our ecosystem can bring unique capabilities to their customers.”
Facebook's parent company Meta says Llama 3.1 is “the world's largest, most capable openly available infrastructure model” and is comparable to commercial products from companies such as: Open AIGoogle, Anthropic, etc. The new version of Llama has 405 billion parameters, or tweakable factors. Meta has already released two smaller versions. Llama 3One has 70 billion parameters and the other has 8 billion parameters. Meta has also released upgraded versions of these models under the brand name Llama 3.1.
“Llama 3.1 405B is the first publicly available model that matches leading AI models in terms of state-of-the-art capabilities in general knowledge, maneuverability, mathematics, tool use and multi-language translation,” Mehta said, sharing details about the AI model.
Unlike OpenAI and Google's latest models, however, Llama is not “multimodal,” meaning it's not built to process images, audio, and video. But Meta says the model is very good at working with other software, such as web browsers. The announcement also highlights Meta's close relationship with Nvidia, a key partner that provides Facebook's parent company with computing chips called GPUs to help train AI models, including the latest version of Llama.
With the release of Llama 3.1, Meta has once again criticized the industry, showing that a closed approach to large-scale AI development might not be the right way to go, but it also finds itself at the center of a debate about the dangers of releasing AI without controls and safeguards.
But the open-source approach that Meta is emphasizing could cost the company billions of dollars, lowering costs because it has vast resources to develop and deploy AI models at scale. Google and OpenAI keep their most advanced models private.
“I believe AI will evolve in a similar way,” Zuckerberg said in the letter. “Today, a few tech companies are developing state-of-the-art closed models, but open source is rapidly closing the gap.”
Experts believe that Meta's open source approach will benefit developers who can get exposure to the company's tools and build applications on top of them, helping to grow the open source developer community.
© IE Online Media Services, Inc.
First uploaded: July 24, 2024, 8:00 AM
