Meta Platforms announced on Saturday the release of its latest large language models, Llama 4 Scout and Llama 4 Maverick, describing them as its “most advanced models yet” and “the best in their class for multimodality.” These models are part of Meta’s push into multimodal AI, systems capable of understanding and generating content across text, images, video, and audio.
Both Llama 4 Scout and Maverick will be available as open-source software, signaling Meta’s continued commitment to collaborative AI development. In addition, the company offered a preview of Llama 4 Behemoth, a powerful upcoming model it calls “one of the smartest LLMs in the world,” designed to serve as a teacher for future systems.
The release follows reports that Meta had delayed the launch of Llama 4 due to the models falling short of internal benchmarks in reasoning and math tasks. Concerns also arose over their performance in humanlike voice conversations compared to competitors like OpenAI.
As part of its broader AI ambitions, Meta plans to invest up to $65 billion this year in AI infrastructure, responding to growing investor pressure and intensifying competition in the AI space sparked by the success of AI tools.
No comments:
Post a Comment