Meta Platforms Inc., the parent company of Facebook, unveiled its latest AI model, Llama 3.1, which CEO Mark Zuckerberg has described as "state of the art." Trained over several months with hundreds of millions of dollars in computing power, Llama 3.1 represents a significant upgrade from Llama 3, released in April. This model powers chatbots both within Meta and for external developers, offering advanced capabilities such as improved reasoning for complex problems, generative AI for image creation, and the innovative "Imagine Yourself" feature. Meta uses Llama to run its AI chatbot, Meta AI, across apps like Instagram and WhatsApp, and it expects the chatbot to become the most widely used globally by year-end.

Zuckerberg emphasized the immense investment Meta has made in AI, with training costs for Llama 3.1 reaching hundreds of millions of dollars. Future models will likely require even more substantial financial outlays. Despite scaling back on some futuristic technologies and management layers in what Zuckerberg called the “year of efficiency,” Meta continues to prioritize AI investments. By making Llama’s technology available for free under an acceptable use policy, Meta aims to position its work as the foundation for future innovations, reducing reliance on competitors. However, the company keeps the data sets used for training Llama 3.1 confidential, though they include publicly available posts from Facebook and Instagram and other licensed proprietary data.