Meta Introduces Llama 3.1: An AI Challenger

Llama 3.1 includes pretrained and instruction-tuned text in/text out open source generative AI models with 8B, 70B, and 405B parameters

Llama 3.1’s debut today shows tremendous progress towards that goal, from dramatically enhanced context length to tool use and multilingual features

Meta and IBM launched the  AI Alliance in December 2023 with over 50 global initial members and collaborators

Llama 3.1 gives the global AI community an open, state-of-the-art model family and development ecosystem to explore, experiment

The release features strong new models, system-level safety safeguards, cyber security evaluation methods, and improved inference-time guardrails

Llama 3.1 improves all model sizes, but the 405B open source model matches leading proprietary, closed source LLMs for the first time

Llama 3.1-405B may be built upon, modified, and run on-premises, unlike its closed source contemporaries