Meta Open Sources LLaMA 2: Democratizing Large Language Models

Meta has released LLaMA 2, a collection of pretrained and fine-tuned large language models (LLMs) ranging in scale from 7 billion to 70 billion parameters. Unlike its predecessor, LLaMA 2 is available for both research and commercial use, marking a significant step toward democratizing access to advanced AI technology. The models were trained on 2 trillion tokens of data and show substantial improvements in helpfulness and safety compared to LLaMA 1. Meta has also released Code Llama, a specialized version for code generation tasks.