Meta poached three core researchers from OpenAI, Zuckerberg's "money ability" proved effective

In the battle for artificial intelligence talent, Zuckerberg's "cash ability" offensive has once again proven effective. Despite OpenAI CEO Sam Altman publicly mocking Meta's high-salary poaching strategy, Meta has successfully recruited three core researchers from OpenAI.

According to the latest report from The Wall Street Journal, Meta recently succeeded in poaching three researchers from OpenAI: Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. These three were previously responsible for establishing OpenAI's Zurich office and have now joined Meta's Superintelligence team.

Public information shows that they collectively joined OpenAI in December 2024 and co-founded the Zurich office, and within less than a year, they have collectively jumped to Meta. An OpenAI spokesperson has confirmed that these three researchers have left.

This latest victory proves that Zuckerberg's widely reported recruitment offensive is yielding results.

An earlier article from Wall Street Journal mentioned that Zuckerberg has adopted an unprecedented recruitment strategy. He is personally assembling a secret AI team called "Superintelligence," aiming to achieve Artificial General Intelligence (AGI). Zuckerberg has even personally contacted hundreds of top AI researchers via WhatsApp, coordinating target candidates through a chat group named "Recruiting Party," and hosting recruitment dinners at his residences in Palo Alto and Lake Tahoe.

Even more surprisingly, Zuckerberg has been offering top talents compensation packages exceeding $100 million in an attempt to lure talent away from OpenAI. Wall Street Journal reported that OpenAI CEO Sam Altman revealed in a podcast: "They started offering huge offers to many people on our team, like $100 million signing bonuses and salaries exceeding that amount per year. It's insane."

Regarding Zuckerberg's charm offensive, Altman stated in the podcast: "I'm really glad that, at least so far, our best talents have not decided to accept his offers." He added, "I think the strategy of offering large upfront guaranteed salaries, and using that as a reason to tell someone to join, really focuses on that rather than on the work and mission, I don't think that will build a great culture."

However, this recruitment strategy has produced mixed results. Zuckerberg recently spent $14 billion to successfully poach Alexandr Wang, CEO of Scale AI, making the 28-year-old one of the most expensive hires in tech industry history.

Yet, larger targets remain unattained. Heavyweights including OpenAI co-founders Ilya Sutskever and John Schulman have not been successfully poached, as both continue to co-found new startups

Llama 4 Major Model Failure, Zuckerberg Deeply Disappointed

Zuckerberg's sense of urgency stems from Meta's recent setbacks in the AI field. Reports indicate that the Llama 4, released in April this year, left Zuckerberg feeling deeply disappointed.

In April, Meta announced the release of the multimodal Llama 4 family, but it quickly faced issues, with developers questioning Meta's premature manual labeling of the test set, making the model appear exceptionally impressive on commonly used rankings.......

The CEO has repeatedly told Meta insiders that he hopes to have the best AI product by the end of the year—both in terms of usage and performance. This goal has forced the AI team to work overtime for an extended period, but the model's performance has still faced internal and external scrutiny. Meta's management believes it has not met expectations, while the developer community criticizes it for "overpromising and underdelivering."

Worse still, Meta's large model "Behemoth" has been delayed; it was originally scheduled to launch in the spring of 2025 but has now been postponed for a long time. Although Meta claims that this model surpasses competing products from OpenAI, Anthropic, and Google, the leadership is concerned that it shows no significant improvement compared to previous models