Publicly available benchmark test results from reputable AI research organizations

Meta Releases Llama 3.1 405B Open-Source AI Model to Rival GPT-4o

Jul 23, 2024, 04:01 PM

Meta has officially released Llama 3.1, its most advanced open-source AI model to date, featuring three variants: 8B, 70B, and 405B parameters. The 405B model, in particular, is being touted as the first open-source model to rival top proprietary models like OpenAI's GPT-4o and Claude Sonnet 3.5. Llama 3.1 excels in various benchmarks, including abstract reasoning, creative ideation, and multilingual capabilities, supporting eight languages. The model also boasts a 128K context length, 15 trillion tokens, and was trained on 16,000 H100 GPUs. New safety tools are also included. Mark Zuckerberg emphasized the significance of open-source AI, suggesting that it is the path forward for the industry. The release marks a pivotal moment in AI, potentially shifting the industry towards open-source solutions.