Meta launches to solidify its position as a leader in open-source AI.

New LLM ‘lama 3.1’ released, ‘Meta, the parent company of Facebook, has open-sourced its latest artificial intelligence (AI) model.’,
,
, ‘Meta announced on the 23rd (local time) the release of a new version of large language model (LLM) called lama 3.1. It includes three types: lama 3.1 405B with 40.5 billion parameters, medium model lama 3.1 70B, and small model lama 3.1 8B.’,
,
, ‘In the training of lama 3.1, NVIDIA’s latest graphics processing units (GPUs) H100 were used. Compared to the previous models, 8 additional languages are supported, and a new feature has been added that generates images that look like actual photos when users describe a person’s appearance. According to Meta, lama 3.1 outperformed ‘OpenAI’s GPT-4o’ and ‘Anthropic’s Claude 3.5 Sonnet’ in ‘Massive Multi-Task Language Understanding (MMLU).”,
,
, ‘Like its previous models, lama 3.1 is also open-source, which sets it apart from competitors such as Google and OpenAI.’,
,
, ‘Silicon Valley=Special correspondent Song Young-chan 0full@hankyung.com\n’]