Meta has unveiled its latest open source large language model family, Llama 2, which comes in various sizes with pre-trained versions available for 7B, 13B, and 70B parameters. The Llama 2-chat variant is fine-tuned for chat-style interactions through supervised fine-tuning and reinforcement learning with human feedback (RLHF). Meta has trained Llama 2 on 2 trillion tokens of data from publicly available sources, increasing its performance on common LLM benchmarks. The model's license allows for research and commercial use, setting a new standard for open source large language models and potentially impacting the AI industry as much as ChatGPT's launch.