Inside Meta’s race to build superintelligence: How TBD Lab is driving the next Llama model and poaching top AI talent

Inside Meta’s race to build superintelligence

Meta Platforms has officially named its newly formed machine superintelligence division “Meta Superintelligence Labs” (MSL) and given a key role to a sub-team called TBD Lab, the Wall Street Journal reported. The unit, led by Chief AI Officer Alexandr Wang—who joined after Meta’s $14 billion stake purchase in his startup Scale AI—is tasked with building a computer mind more capable than any human. CEO Mark Zuckerberg has described the mission as creating “personal superintelligence for everyone.”

TBD Lab takes charge of Llama’s evolution

At the heart of the push is TBD Lab, short for “to be determined,” which is spearheading development of the next version of Llama, Meta’s large language model. The project is led by Jack Rae, a former Google researcher. Internally referred to as “Llama 4.5” or “Llama 4.X,” the model aims to extend reasoning abilities and support advanced AI agents. Wang said in a recent memo that the team is collaborating with other AI units to “parallelize across several efforts” and reach frontier results faster.

A hiring spree targeting top AI talent

To fuel its ambitions, Meta has launched an aggressive recruitment campaign, poaching at least 18 researchers from OpenAI and several from Google, including recent hires Tong He and Yuanzhong Xu. Pay packages have reached extraordinary levels, with some offers in the hundreds of millions of dollars and at least two cases reportedly at $1 billion over several years. Nine staff members from Meta’s internal infrastructure team were also shifted to TBD Lab after receiving job offers from Thinking Machines Lab, a startup run by former OpenAI executive Mira Murati.

Zuckerberg’s vision for superintelligence

Zuckerberg has framed the superintelligence initiative as a transformative leap for technology, saying it will “improve all our existing systems and enable the creation and discovery of new things that aren’t imaginable today.” The company hopes that combining deep research talent with massive compute resources will accelerate progress toward models with advanced reasoning and real-world task capabilities.

A strategic shift in Meta’s AI organization

The formation of MSL consolidates all of Meta’s AI initiatives under one umbrella, allowing greater coordination across infrastructure, research, and product integration. Wang’s leadership marks a pivot toward scaling up both technical capacity and recruitment speed. With TBD Lab at the center, Meta is aiming to outpace rivals in the race toward superintelligence—while continuing to refine and deploy its Llama models across consumer and enterprise applications.