On Tuesday, Meta announced the release of Llama 3.1, the latest version of its big language model, which it claims currently rivals rivals OpenAI and Anthropic. The new model comes just three months after Meta launched Llama 3, integrating it into a chatbot called Meta AI. he lives now On Facebook, Messenger, Instagram and WhatsApp and but also powers the company’s smart glasses. in between, OpenAI and anthropic have already released new versions of their AI models, a sign that Silicon Valley’s AI arms race isn’t slowing down anytime soon.
Meta he said The new model, called Llama 3.1 405B, is the first outdoor model that can compete with competitors in general knowledge, math skills and translation in multiple languages. The model has been trained on more than 16,000 NVIDIA H100 graphics processorsthe fastest available chips, which currently cost about $25,000 each and can beat competitors in more than 150 benchmarks, Meta claimed.
“405B” refers to the 405 billion parameters, the internal variables that the AI model uses to reason and make decisions. The greater the number of parameters of an AI model, the smarter we perceive it to be. OpenAI’s GPT-4 model, in comparison, is reported to be About 1.5 trillion parameters, although the company has not disclosed this number to date. In addition, Meta also released improved versions of the existing Llama models with 70 billion and 8 billion parameters each, and claimed that the new versions have, among other things, stronger reasoning capabilities.
Developers can download Llama 3.1 from it official sitewhereas regular users can play with it through Meta AI on or using WhatsApp meta.ai, the company’s website for the chatbot. “Llama 405B’s advanced reasoning capabilities enable Meta AI to understand and answer your more complex questions, especially in math and coding,” Meta’s blog post says. “Get help with your math homework with step-by-step explanations and feedback, and code faster with debugging support and optimization.” (Editor’s note: Engadget will oppose Llama 3.1 New York Times Spelling Bee and let you know). For now, Meta AI on Facebook, Messenger, and Instagram is still limited to the minor version of Llama 3.1, which uses 70 billion settings.
Unlike OpenAI, Google, Microsoft, and Anthropic, which keep their AI models proprietary, Meta’s AI models are open source, meaning anyone can modify and use them for free without sharing their personal data with Meta. One letter Meta CEO Mark Zuckerberg, released on Tuesday, argued that an open-source approach to the development of artificial intelligence will provide greater access to the benefits of the technology, prevent the concentration of power among a few large companies and allow for safer AI implementation in society. By open-sourcing the company’s largest language model to date, Meta aims to make Llama the “industry standard” for developing AI-powered apps and services for everyone, Zuckerberg writes.
Open-source AI models and adding them to its existing products used by billions of people could allow Meta to compete more effectively with OpenAI, which is fueling the AI explosion when ChatGPT and DALL-E chatbots launch in 2022. And it can increase engagement — Meta announced today users will soon be able to add AI-generated images directly to feeds, stories, comments and messages on Facebook, Messenger, WhatsApp and Instagram.
In his letter, Zuckerberg also criticized Apple and its closed ecosystem, arguing that the iPhone maker’s restrictive and arbitrary policies limit what Meta can build on its platforms. “[It’s] If we can build the best versions of our manufacturers and not limit what our competitors can build, it’s clear that Meta and many other companies will be freed up to build better services for people.”
