Llama 4's Flop Reportedly Forced Zuckerberg to 'Handpick' Meta's New AI Team
Disappointed by the latest open-source large model Llama 4, Meta Platforms founder Mark Zuckerberg is personally assembling a new team of AI experts in an effort to surpass all competitors and ultimat
Disappointed by the latest open-source large model Llama 4, Meta Platforms founder Mark Zuckerberg is personally assembling a new team of AI experts in an effort to surpass all competitors and ultimately achieve the so-called "Artificial General Intelligence" (AGI).
Insiders reveal that Zuckerberg has prioritized the recruitment of this secretive new team, internally referred to as the "Superintelligence Group." He has also created an executive group chat named "Recruiting Party," where members discuss talent scouting around the clock.
Media reports indicate that Meta has offered compensation packages in the "seven- to nine-figure" range to researchers from renowned AI companies like OpenAI and Google, with some already agreeing to join.
It is reported that the founder's sudden hands-on approach is partly due to his deep disappointment with the latest Llama 4 model released in April.
He had previously stated multiple times within Meta that the company would launch AI products with top-tier applications and performance by the end of this year. This demand has placed immense pressure on the AI team, forcing employees to work around the clock. However, the final delivered model faced criticism both internally and externally - company executives deemed its performance unsatisfactory, while external developers accused the company of overpromising and underdelivering, even exaggerating the capabilities of some products.
It is said that Zuckerberg's anger runs "even deeper" than that of external developers - he is furious that outsiders believe he attempted to 'cover-up' the underwhelming performance of the latest product.
Subsequently, Meta delayed the release of its largest-parameter "Behemoth" model, which it claimed would outperform competitors like OpenAI and Google. Reports suggest that company executives were concerned the model's actual improvements over its predecessor were limited, despite the high-profile marketing campaign.
Insiders say witnessing the failure of Llama 4 has fueled Zuckerberg's interest in forming a new team. He has been personally drafting the recruitment list, involving himself in almost every candidate's hiring process, and preferring to act as the first point of contact during negotiations. He has even rearranged the seating layout at Meta's headquarters to ensure new hires can sit near him.
Reports indicate that Zuckerberg plans to hire around 50 people for this new team, including a new head of AI research.
While building this team, Meta is also planning a multi-billion-dollar investment in data labeling and cleaning company Scale AI, which would mark the company's largest external investment to date. Upon completion of the deal, 28-year-old Scale AI founder Alexandr Wang is expected to join Zuckerberg's superintelligence team.
Insiders reveal that Zuckerberg harbors an ambitious goal. In his view, Meta has the capability - and the responsibility - to surpass other tech companies and achieve AGI, where machines can perform at human levels across multiple tasks. Once the development team reaches this milestone, the AI capabilities could be integrated into Meta's products, such as its social apps and AI-powered smart glasses.
Zuckerberg also believes the company's advertising business is strong enough to fund the AI competition at a scale of tens of billions of dollars. He has previously promised to invest "hundreds of billions of dollars" in AI over the coming years.
Over the past month, Zuckerberg has frequently invited AI researchers, infrastructure engineers, and other entrepreneurs to his California residence to discuss joining Meta. According to those familiar with the discussions, he has told potential recruits that Meta has sufficient cash flow to build a "multi-gigawatt" data center, giving the company one of the most powerful computing infrastructures in the world.
Disclaimer: The views in this article are from the original Creator and do not represent the views or position of Hawk Insight. The content of the article is for reference, communication and learning only, and does not constitute investment advice. If it involves copyright issues, please contact us for deletion.