Published AI benchmark results from reputable AI research blogs or Meta's official announcements.

Meta Launches Llama 3 AI Model: 8B, 70B, and Training 400B Versions

Apr 18, 2024, 04:13 PM

Meta Platforms has officially released its latest generative AI model, Llama 3, which includes both 8B and 70B parameter versions, instruction-tuned, and trained on a dataset of 15 trillion tokens. These models offer enhanced language understanding and task execution across various applications. The release also includes a 400B parameter model still in training, promising to surpass the capabilities of models like Claude 3 Opus. Additionally, Meta's AI chatbot, powered by Llama 3, has been integrated across its platforms, and the Meta AI virtual assistant will be updated to include these new models. Llama 3 supports an 8k context length, contributing significantly to the open-source community.