Meta Set to Launch First AI Models Built Under Alexandr Wang’s Leadership

Meta is preparing to unveil its first AI models developed under its Superintelligence group, led by Alexandr Wang. According to Axios, the company plans to eventually release open-source versions of these models, marking a significant milestone in Meta’s aggressive push toward artificial general intelligence.

Meta established its Superintelligence project in June 2025, with CEO Mark Zuckerberg expressing ambitions to bring “personal superintelligence to everyone.” Wang joined the initiative after Meta acquired his company, Scale AI, for $15 billion, and he now leads the group’s efforts.

A Race to Close the Benchmark Gap

The upcoming release carries considerable weight. Meta’s previous Llama model release fell noticeably behind competitors in industry benchmarks, making these new models a critical opportunity for the company to reassert itself. Meta has poured hundreds of billions of dollars into AI development, positioning artificial intelligence as central to its long-term strategy.

Experts Remain Divided on the Approach

Not everyone is convinced the current path leads to true machine intelligence. Yann LeCun, Meta’s long-serving AI chief who departed last year, has publicly argued that today’s generative AI models — which essentially match queries to likely responses using vast datasets — are not genuinely thinking systems. LeCun believes a fundamentally different approach is needed to reach human-level AI, though he remains cautiously optimistic that such a breakthrough could arrive within the next decade through open-source collaboration and focused research.

Where Meta’s Superintelligence models fit within this broader debate will be closely watched across the industry.