Meta has hired influential OpenAI researcher Trapit Bansal to work on artificial intelligence reasoning models as part of the company’s new superintelligence division, a person familiar with the matter told TechCrunch.
OpenAI spokesperson Kayla Wood confirmed to TechCrunch that Bansal has left OpenAI. Bansal’s LinkedIn page states that he left OpenAI in June.
Bansal has been with OpenAI since 2022 and was a key player in launching the company’s reinforcement learning work with co-founder Ilya Sutskever. He is one of the founders of OpenAI’s first AI reasoning model, o1.
Bansal can significantly strengthen Meta’s superintelligence lab, which also includes leaders such as former Scale AI CEO Alexander Wang, as well as former GitHub CEO Nat Friedman and Safe Superintelligence co-founder Daniel Gross. Bansal can help Meta develop an advanced AI reasoning model that will be competitive with leading technologies such as OpenAI’s o3 or DeepSeek’s R1. Currently, Meta does not offer an AI reasoning model in the public domain.
In recent months, Mark Zuckerberg has been actively recruiting to build a new AI team at Meta, offering compensation packages of $100 million to the best researchers who join his company. It is unclear what exactly was offered to Bansal in this deal.
Nevertheless, it seems that Zuckerberg managed to lure away the best AI researchers.
Three other former OpenAI researchers – Lucas Bayer, Alexander Kolesnikov, and Xiaohua Zhai – have also joined the Meta superintelligence team in recent weeks, The Wall Street Journal reported on Wednesday. Bansal will join them along with former Google DeepMind researcher Jack Ray and former Sesame Street machine learning leader Johan Shalkvik, Bloomberg reports.
To fill out his new AI division, Zuckerberg reportedly tried to acquire startups with strong AI research labs, such as Sutskever’s Safe Superintelligence, Mira Murati’s Thinking Machines Labs, and Perplexity. However, these negotiations never reached the final stage.
In a recent podcast, OpenAI CEO Sam Altman said that Meta tried to poach his startup’s top talent, but claimed that “none of our best people decided to take the offer.”
AI reasoning models are a key area for Meta’s superintelligence team to succeed. Over the past year, OpenAI, Google, and DeepSeek have released high-performance AI reasoning models that have pushed the boundaries of software capabilities. By teaching AI models to work through problems before giving an answer, using additional time and computing resources, AI labs have succeeded in improving AI performance in tests and real-world tasks.
Meta’s AI lab could become a key internal group that powers products across the company, similar to Google’s DeepMind division. Under the leadership of Clara Shih, former general manager of AI at Salesforce, Meta has ambitious plans to create AI agents for businesses. In order to create competitive agents, Meta must develop advanced AI reasoning models to power them.
With the addition of Bansal and other key AI researchers, Meta hopes to pull ahead in the AI race. This could be challenging given that OpenAI plans to release an open AI reasoning model in the coming weeks – a proposal that could put even more pressure on Meta’s open AI offerings.
