Mark Zuckerberg has quietly brought together about 50 of Meta’s top AI researchers and engineers into a new “Superintelligence” squad. Their mission: build systems that can rival—or even surpass—human reasoning. By branding this effort so explicitly, Meta is signaling just how seriously it takes the race for artificial general intelligence.
A $65 Billion AI Push and the Future of Llama 4
Next year’s budget includes a massive $65 billion AI investment, part of which will go toward a sprawling data center. The urgency behind this cash infusion comes after Llama 4 fell short of performance goals. In response, Zuckerberg has redirected resources to an even bolder project—codenamed “Behemoth”—though its debut has now slipped into the fall. The debate inside Meta centers on whether supercharging infrastructure and hiring fresh talent will outpace the returns from fine-tuning existing models.
Behemoth—Meta’s Next Flagship Model
Details remain scarce, but one thing is clear: executives are willing to endure delays for a more polished product. That age-old tension—launch quickly or wait to wow users—could determine whether Meta resets expectations for commercial AGI prototypes.
High-Profile Hires and Cutthroat Competition
A standout recruit is Alexandr Wang, founder of Scale AI, who’s set to join once Meta finalizes a multibillion-dollar backing of his startup. It’s a direct challenge to rivals like Microsoft, Google and OpenAI, all of whom have pulled ahead with advanced chatbots and major partnerships. Zuckerberg’s hands-on recruitment push reflects his belief that Meta’s vast data stores, engineering talent and computing resources still give it a unique edge.
Meta’s Strategy in the Broader AGI Race
With competitors from OpenAI to Elon Musk’s xAI scrambling for leadership, Meta’s aggressive talent drive and huge investments underline how high the stakes have become. In the coming year, we’ll see whether the winning formula relies on deep pockets, concentrated talent teams—or some hybrid approach no one’s tried yet.