Meta releases Llama 3 open LLMs (8B/70B) with Llama Guard 2 and Hugging Face integration

AI Impact Summary

Meta's Llama 3 release expands open-access LLM options with 8B and 70B variants, including an 8B model using Grouped-Query Attention and a new 128,256-token tokenizer. The upgrade affects deployment characteristics through bigger embeddings and longer-context potential, and Llama Guard 2 adds a safety-focused fine-tune that will influence content moderation workflows. By aligning with Hugging Face and Transformers, the release lowers integration friction for teams, but attribution requirements and derivative-name rules will impact deployment branding and licensing compliance.