Meta has hired Trapit Bansal, a prominent AI researcher who previously worked at OpenAI, to join its newly formed AI superintelligence team. According to TechCrunch, Bansal will focus on building advanced AI reasoning models—an area increasingly central to Meta’s long-term AI ambitions. Bansal, who joined OpenAI in 2022, played a key role in the company’s reinforcement learning research and collaborated closely with co-founder Ilya Sutskever. He was also among the original contributors to OpenAI’s first reasoning model, known as o1. His LinkedIn profile shows that he left OpenAI in June, and the departure was confirmed by spokesperson Kayla Wood.

His move is seen as a significant addition to Meta’s new AI unit, which already includes notable figures like former Scale AI CEO Alexandr Wang. The team is reportedly in discussions with other influential names such as Nat Friedman and Daniel Gross. With Bansal’s expertise, Meta aims to catch up with rivals like OpenAI and DeepSeek, which have already released sophisticated AI reasoning models such as o3 and R1. As of now, Meta has yet to unveil a comparable offering.

Mark Zuckerberg has been actively pursuing top AI talent in recent months, reportedly offering compensation packages reaching $100 million. While details of Bansal’s offer remain undisclosed, his decision to join reflects the effectiveness of Meta’s recruitment efforts. According to The Wall Street Journal, Bansal joins several other recent hires from OpenAI, including Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. Bloomberg also reports that former DeepMind researcher Jack Rae and Johan Schalkwyk from startup Sesame are now part of Meta’s team.

Meta had also explored acquiring AI startups like Safe Superintelligence, Thinking Machines Labs, and Perplexity to enhance its capabilities, though none of these deals materialized. In a recent podcast, OpenAI CEO Sam Altman responded to Meta’s recruitment push, saying, “None of our best people have decided to take him up on that.”

The development of advanced reasoning models has become a central focus for Meta, as these systems are essential for handling complex tasks through multi-step logical thinking. This approach, which often requires more computational resources, has led to better performance in both testing and real-world use. Meta’s AI superintelligence team is expected to become a vital component of the company’s broader AI vision, similar to how DeepMind functions within Google. The team’s future goals include creating enterprise-grade AI agents, with leadership from Clara Shih, the former AI CEO at Salesforce.