TLDRs;
- Meta CEO Mark Zuckerberg is personally leading a new AGI-focused team known internally as “Superintelligence”
- The team follows underwhelming performance of Meta’s Llama 4 and delays in launching the next-generation model, Behemoth
- Meta is reportedly in talks for a $10 billion investment in Scale AI to bolster its data infrastructure for model training
- Authors and researchers remain divided over Meta’s definition and direction in the race toward artificial general intelligence
Meta Platforms is making a high-stakes pivot toward artificial general intelligence, with CEO Mark Zuckerberg now taking the reins of a newly formed team internally branded as “Superintelligence.”
The initiative is designed to push Meta into the upper echelon of AGI research and development, at a time when the company’s current flagship AI models have not met expectations.
Meta’s Big Bet on AGI
Zuckerberg, long known for his hands-on approach during moments of strategic reinvention, has reportedly met face-to-face with candidates for the Superintelligence team, hosting private recruitment talks at his homes in both Palo Alto and Lake Tahoe. The team will consist of roughly 50 top-tier AI researchers and engineers, including a newly appointed head of AI research, and will be stationed near Zuckerberg’s personal office at Meta’s Menlo Park headquarters. This move signals the CEO’s intent to stay deeply embedded in the direction and performance of Meta’s next chapter in artificial intelligence.
Llama 4 Disappointment Sparks Urgency in AI Overhaul
The launch of Meta’s Llama 4 model earlier this year was expected to position the company competitively against the likes of OpenAI and Google. Instead, the model received a lukewarm response, with developers citing limited capabilities compared to other cutting-edge systems. Compounding these concerns are internal delays in releasing the much-anticipated Behemoth model, a project that aims to deliver larger-scale performance but has yet to see the light of day.
The underwhelming reception of Llama 4 appears to have accelerated Meta’s internal restructuring of its AI efforts. The formation of a dedicated AGI team marks a break from the company’s earlier research structure under FAIR, which prioritized broad technical innovation over narrow, mission-driven AGI development. Now, the focus has shifted to building models that could someday rival human-level intelligence, with a strong emphasis on direct leadership and resource concentration.
$10 Billion Investment in Scale AI
Meta’s ambition does not stop at talent acquisition. The company is currently in talks to invest over $10 billion in Scale AI, a data infrastructure startup specializing in high-quality labeling services. If finalized, the investment would become Meta’s largest external AI deal to date. The partnership is expected to significantly boost the company’s ability to train increasingly complex models by tapping into Scale AI’s platform, which spans thousands of global data contributors.
Unlike some of its AI competitors who rely heavily on outside fundraising, Meta is financing this expansion through its own reserves, largely supported by its advertising business. This self-funded model gives the company greater autonomy as it pursues a long-term vision for AGI, while others in the industry remain constrained by investor timelines and revenue targets.
That said, while Meta’s push into superintelligence has generated buzz, it also raises questions about how the company defines AGI and what it hopes to achieve in the short term. Industry experts remain divided on whether current AI approaches are capable of achieving human-level cognition. For Meta, the goal may be less about reaching philosophical milestones and more about embedding smarter AI into its consumer ecosystem, from social media assistants to augmented reality glasses.
