Frustrated by AI Setbacks, Zuckerberg Personally Recruits Top Talent to Drive Meta’s Ambitious Superintelligence Quest — Can This New Team Close the Gap?

Mark Zuckerberg Takes Matters Into His Own Hands to Build a New AI Dream Team

Mark Zuckerberg is taking a more direct role in building Meta’s next generation of AI experts.

Frustrated by the company’s recent AI setbacks, the Meta CEO has been personally reaching out to researchers and engineers, inviting them to join what insiders call a “superintelligence group.”

This roughly 50-strong team is tasked with pushing Meta beyond current AI capabilities, aiming to develop artificial general intelligence (AGI) — AI with human-level thinking — and ultimately superintelligence, which would surpass human intelligence.

What’s Behind Meta’s Urgent AI Team Expansion

Sources familiar with Meta’s plans reveal Zuckerberg’s dissatisfaction with the reception of Llama 4, Meta’s latest large language model.

Despite significant investment, the model struggled to outshine competitors on public leaderboards like LMArena, where Google’s Gemini 2.5 Pro and OpenAI’s o4 hold the top spots.

The controversy deepened when Meta was found to have submitted a non-public version of Llama 4 to the leaderboard, sparking criticism.

The setback has prompted Zuckerberg to shift into what has been described as “founder mode,” with the CEO personally managing recruitment efforts.

He has hosted dinners and lunches, and even created a WhatsApp group called “Recruiting Party” where senior Meta executives collaborate on attracting top talent.

Venture capitalist Deedy Das commented on the recent flurry of AI researchers moving from Meta to rivals like OpenAI and Anthropic, despite high salaries offered, stating,

“The AI talent wars are absolutely ridiculous.”

What Does The Superintelligence Group Aim To Achieve

The newly formed team will focus on improving Llama’s performance and developing fresh AI tools for integration into Meta’s core products, including Facebook, Instagram, WhatsApp, and AI-enhanced Ray-Ban smart glasses.

Meta has been revitalising its AI glasses project, aiming to embed facial recognition and virtual assistant features.

This effort puts it in direct competition with Google’s AI-powered XR glasses and OpenAI’s work on omnipresent AI companions.

How Big Is Meta’s AI Investment In 2025

Earlier this year, Zuckerberg announced a $65 billion investment plan for AI-related projects, including a new large data centre.

Meta also revealed plans to raise capital expenditures to between $64 billion and $72 billion this year to boost AI capacity.

These figures reflect a significant bet on AI’s role in Meta’s future.

What Happened To Meta’s Flagship AI Model Behemoth

Meta has delayed the release of its new flagship AI model, known as “Behemoth.”

Originally expected in April, the launch has now been pushed back to later in the year due to questions over whether the model’s improvements over Llama 4 are substantial enough.

The Wall Street Journal reported that internal challenges in training Behemoth have slowed progress.

Previously, Meta had suggested Behemoth was ahead of rival models from OpenAI, Google, and Anthropic, but these claims appear to have been reassessed.

Who Is Joining The Team And What Are The Offers

Among those reportedly joining the superintelligence team is Alexandr Wang, CEO of Scale AI, an AI startup valued at nearly $14 billion.

Meta is reportedly negotiating a $10 billion investment in Scale AI, which would be its largest external investment yet.

Wang and several Scale AI employees are expected to move to Meta following the deal’s closure.

Compensation packages for new recruits are said to range from seven to nine figures, reflecting the fierce competition for AI talent.

Zuckerberg’s personal involvement extends to reorganising seating at Meta’s Menlo Park headquarters, placing the new team near him to foster closer collaboration.

How Is Meta’s AI Performance Viewed Against Rivals

Meta was seen as trailing Microsoft, Google, and OpenAI in the race to develop advanced AI.

However, the company’s recent spending surge has helped narrow that gap.

Zuckerberg claimed earlier in 2025 that over one billion users interact with Meta AI monthly across its apps, underscoring the company’s reach despite the challenges.

Meanwhile, other AI players, including OpenAI and Elon Musk’s xAI, are also intensifying efforts to develop AGI, securing additional funding to stay competitive in what has become a high-stakes race.

Meta shares have gained around 17% year-to-date, reflecting investor confidence despite the hurdles.

Zuckerberg’s hands-on approach to assembling this “superintelligence group” reveals the pressure the social media giant faces to maintain relevance in the fast-moving AI landscape.