Meta has officially launched Llama 3.1, its largest open-source AI model with 405 billion parameters, which claims to outperform OpenAI's GPT-4o and Anthropic's Claude 3.5 Sonnet on various benchmarks. CEO Mark Zuckerberg anticipates that this AI assistant will become the most widely used by the end of the year, surpassing ChatGPT, and is expanding its availability across multiple languages and platforms including Facebook, Instagram, and WhatsApp. Trained on over 16,000 NVIDIA H100 GPUs, Llama 3.1 is designed to be cost-effective, with operational expenses approximately half of those associated with GPT-4o. Meta's open-source strategy aims to foster a competitive ecosystem while addressing concerns about potential misuse of such technology. Despite the significant investment in development, Meta's approach is rooted in the belief that open-source models will evolve more rapidly than proprietary systems. Collaboration with major companies like Microsoft and Google is set to enhance the deployment of Llama 3.1 across various applications.