Meta hires key OpenAI researcher to work on AI reasoning models

In a significant move to bolster its AI capabilities, Meta has recruited Trapit Bansal, a prominent researcher from OpenAI, to enhance its AI reasoning models. This transition comes as Meta launches its new AI superintelligence division, according to sources familiar with the situation. Confirming Bansal's departure from OpenAI, spokesperson Kayla Wood shared that he left the company in June. Bansal, who joined OpenAI in 2022, played an instrumental role in the development of reinforcement learning alongside co-founder Ilya Sutskever and is recognized as a foundational contributor to OpenAI's initial AI reasoning model, o1. His expertise is expected to provide a substantial boost to Meta's AI superintelligence lab, which includes notable figures such as former Scale AI CEO Alexandr Wang. The team is also in discussions with other high-profile recruits, including ex-GitHub CEO Nat Friedman and Safe Superintelligence co-founder Daniel Gross. Meta is keen to develop a competitive frontier AI reasoning model that could rival leading technologies like OpenAI's o3 and DeepSeek's R1. Currently, Meta does not have a publicly available AI reasoning model, which underscores the urgency of this recruitment. Recently, CEO Mark Zuckerberg has aggressively expanded Meta's AI workforce, reportedly offering lucrative compensation packages amounting to $100 million to attract top-tier researchers. While the specifics of Bansal’s compensation remain undisclosed, it is evident that Meta is successfully attracting elite AI talent. In addition to Bansal, three other former OpenAI researchers—Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—have recently joined Meta's AI superintelligence team, as reported by The Wall Street Journal. Bansal will be joined by former Google DeepMind researcher Jack Rae and Johan Schalkwyk, who previously led machine learning efforts at the startup Sesame. To expand its AI unit further, Zuckerberg has explored acquiring startups with strong AI research teams, including Safe Superintelligence and Thinking Machines Labs, but those negotiations did not reach a conclusive agreement. OpenAI CEO Sam Altman revealed in a recent podcast that Meta has actively sought to lure talent from his organization, although he believes that their top researchers have largely remained loyal. A Meta representative chose not to comment on these developments. AI reasoning models are critical for Meta’s ambitions, especially as competitors like OpenAI, Google, and DeepSeek have released advanced models that significantly enhance software capabilities. By training AI to solve problems before providing answers, these labs have achieved remarkable performance improvements on benchmarks and real-world applications. As Meta's AI superintelligence lab aims to become a pivotal force in driving company-wide products, the successful integration of innovative AI reasoning models is essential. With Bansal and other esteemed researchers on board, Meta is poised to make significant strides in the AI landscape. However, the impending release of OpenAI's open AI reasoning model could intensify competition and pressure Meta to accelerate its developments.