Meta's new superintelligence team to spearhead new Llama large-language model, sources say -- WSJ

Meta's new superintelligence team to spearhead new Llama large-language model, sources say -- WSJ

Meta Platforms, Inc. (Meta) has made significant strides in its pursuit of artificial superintelligence with the formation of its new superintelligence team, Meta Superintelligence Labs (MSL). According to recent reports, the team, led by Chief AI Officer Alexandr Wang, is already making notable progress in developing advanced AI models, including the latest iteration of Meta's large language model, Llama.

Initial Progress and Team Composition

Just one month into its operations, MSL has an initial project underway, dubbed TBD Lab, which is spearheading the development of the newest version of Llama. Jack Rae, a hire from Alphabet's Google, is leading this effort. The team is composed of researchers poached from rival AI labs, with some receiving pay packages of tens or hundreds of millions of dollars. This aggressive hiring strategy is part of Meta's broader ambition to build a computer mind smarter than any human's [1].

Mission and Goals

The mission of MSL, as outlined by Chief Executive Mark Zuckerberg, is to "bring personal superintelligence to everyone." The team is working on a variety of projects, including enhancing the reasoning capabilities of AI models and developing AI agents. Alexandr Wang has noted that the team has already made meaningful progress in these areas, enabling more ambitious and parallel efforts [1].

New Llama Model

The new Llama project, currently known internally as Llama 4.5 or Llama 4.X, is being led by Jack Rae and involves collaboration between members of Meta's existing Llama team and TBD Lab. Zuckerberg has described superintelligence as a revolutionary technology that will improve existing systems and enable the creation of new things unimaginable today [1].

Competitive Landscape

Meta's efforts come amidst a competitive landscape where rivals like OpenAI and Google are also making significant strides in AI development. OpenAI, the developer of ChatGPT, recently launched two freely available AI models that can be customized by users. Google, meanwhile, has outlined its latest step toward artificial general intelligence (AGI) with a new model that allows AI systems to interact with convincing simulations of the real world [3].

Investment and Strategy

Meta's aggressive hiring and investment strategy is backed by substantial capital expenditures. In Q2 2025, the company reported $17.0 billion in capital expenditures, primarily targeting AI infrastructure. Zuckerberg's public letter on the vision for personal superintelligence underscores Meta's commitment to empowering individual pursuits rather than automating societal functions through centralized control [4].

Conclusion

Meta's new superintelligence team is poised to make significant advancements in AI technology, with the latest Llama model being a key focus. The company's aggressive hiring strategy and substantial investment in AI infrastructure position it as a key player in the competitive AI landscape. As the team continues to make progress, investors and financial professionals will be closely watching Meta's developments in this critical area.

References

[1] https://www.wsj.com/tech/ai/meta-ai-superintelligence-team-6415a4f4
[2] https://www.aol.com/news/metas-zuckerberg-hiring-ai-team-070056895.html
[3] https://www.theguardian.com/technology/2025/aug/05/openai-meta-launching-free-customisable-ai-models
[4] https://ppc.land/meta-ceo-unveils-personal-superintelligence-vision-for-everyone/