Meta’s frontier AI research team has about 100 people, according to CEO Mark Zuckerberg—a small core inside an AI organization with around 3,400 employees. An infrastructure team of roughly the same size supplies those researchers with compute, pairing a compact research group with equally substantial technical support.
Key takeaways
- Zuckerberg puts Meta’s frontier research headcount at roughly 100.
- A similarly sized infrastructure group supports the researchers.
- Meta Superintelligence Labs employs around 3,400 people overall.
- The research core has roughly doubled from its initial size.
Crypto Briefing reported Zuckerberg’s figures on October 13, 2026, citing a recent video appearance. The research team’s size falls within the range he previously described as ideal for frontier AI work: 50 to 100 researchers.
Meta’s frontier AI research team stays within Zuckerberg’s preferred range
The core research group has grown from approximately 50 people to about 100, roughly doubling while remaining within Zuckerberg’s preferred size range. His stated rationale favors focus over a larger organizational structure.
During a podcast appearance with Rowan Cheung in September 2025, Zuckerberg argued that a small team gives researchers room to concentrate on difficult problems without bureaucracy or rigid deadlines interfering. He described the working dynamic as a “group science project.”
The group sits within Meta Superintelligence Labs (MSL), which Meta established in mid-2025 to bring its separate AI efforts together. The broader organization now employs around 3,400 people—not just the frontier researchers.
Chief AI Officer Alexandr Wang leads MSL, with Nat Friedman serving as co-lead.
An infrastructure team matches the research headcount
Meta maintains an infrastructure group of about 100 people to provide compute for its frontier researchers. The reported headcounts amount to roughly one infrastructure engineer per researcher.
That support encompasses the hardware clusters, training pipelines and systems researchers use to test ideas. In an effort to narrow what it sees as a gap with rival labs, Meta has also announced additional upgrades to its computing capacity.
The infrastructure group is distinct from Meta’s frontier AI research team, even though Zuckerberg described both as approximately the same size.
Recruitment sits at the center of Meta’s AI rebuild
Meta’s restructuring followed a mixed reception for its earlier Llama models and came with an aggressive effort to recruit leading researchers. The company invested heavily in talent and infrastructure throughout 2025.
Meta sought researchers from leading AI firms through exceptional compensation packages. Alongside bringing its existing AI teams under MSL, it set a recruitment goal of an additional 50 to 100 top-tier researchers.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.