Meta is fundamentally reshaping its approach to artificial intelligence through an aggressive Meta AI hiring spree and the creation of specialized research units . Facing intense competition from industry rivals, the technology giant is deploying unprecedented compensation packages and high-profile leadership changes to advance its open-source AI models, reasoning capabilities, and social media algorithms .
The company’s Meta AI hiring spree relies heavily on poaching leading AI experts from competitors like OpenAI, Google, Apple, and TikTok . By offering massive financial incentives and restructuring its internal teams, Meta aims to catch up after its previous models lagged behind the industry standard, while simultaneously supercharging its core advertising business .
Unprecedented Compensation in the War for AI Talent
To staff its newly formed divisions, Meta has upended traditional corporate pay structures . The company is offering compensation packages worth up to $300 million over four years to secure top-tier researchers . These offers feature substantial cash bonuses, highly liquid compensation that exceeds typical stock options, and replacements for equity forfeited by leaving other firms .
Chief Executive Officer Mark Zuckerberg has taken a highly personal approach to recruitment . According to reports, Zuckerberg has made house calls to prospective hires, sometimes bringing homemade soup to convince them to join the company .
This aggressive push has yielded significant results, pulling essential personnel from major competitors . Meta successfully recruited OpenAI researchers Jason Wei and Hyung Won Chung, who previously worked on reasoning models . The company also secured Ruoming Pang, a former supervisor of AI models at Apple, offering a compensation package worth hundreds of millions of dollars over several years . Apple declined to counter the offer, which exceeded the pay of its own top executives outside the CEO role .
In another high-profile acquisition, Meta hired Andrew Tulloch, a co-founder of Thinking Machines Lab . After initially turning down a package that included bonuses worth $1.5 billion, Tulloch changed his mind a few months later and joined the company . Meta also recruited three other researchers from the same $12 billion startup .
Building the Meta Superintelligence Labs
A central focus of Meta’s hiring spree is the Meta Superintelligence Labs, an ambitious unit announced in the summer of 2025 . Led by former Scale AI Chief Executive Officer Alexandr Wang, the division aims to develop industry-leading models .
Recently, the unit hired Trapit Bansal, a highly influential researcher who departed OpenAI in June . Bansal was a foundational contributor to OpenAI’s first reasoning model and a key figure in its reinforcement learning efforts . His expertise is expected to help Meta develop frontier AI reasoning models capable of competing with technologies like OpenAI’s o3 and DeepSeek’s R1 . Currently, Meta does not offer an AI reasoning model to the public .
Bansal joins several other recent additions to the superintelligence team, including former OpenAI researchers Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai . The unit has also added former Google DeepMind researcher Jack Rae and Johan Schalkwyk, a former machine learning leader at the startup Sesame .
Despite these notable departures from OpenAI, its CEO Sam Altman recently claimed on a podcast that Meta had been trying to poach top talent but that “none of our best people have decided to take him up on that” . A Meta spokesperson declined to comment on the matter .
Beyond individual hires, Zuckerberg reportedly attempted to acquire entire startups boasting heavy-hitting research labs, including Safe Superintelligence, Thinking Machines Labs, and Perplexity . However, those discussions did not progress to a final stage .
Forming an Elite Algorithm Research Unit
While the superintelligence lab focuses on frontier models, Meta has also reorganized its internal teams to maximize revenue from its existing platforms . Last October, the company quietly formed an elite unit called MRS Research .
Operating within the Meta Recommendation Systems division, MRS Research is tasked with optimizing the algorithms that power the home feeds of Facebook and Instagram . By working closely with the company’s advertising division, the unit aims to leapfrog current recommendation systems and use AI to superpower Meta’s core business . In late 2025, Meta launched a new AI model that it reported increased ad performance by serving more relevant content to users .
The unit is overseen by Yang Song, the vice president of recommendation research . Song joined Meta in November 2025 after leading user growth and recommendations at TikTok . His team has rapidly expanded, adding prominent experts such as Amazon AI researcher Lihong Li, former OpenAI researcher Xiaolong Wang, and Google researcher Fei Sha .
A Hybrid Approach to Open-Source Models
As Meta absorbs this influx of talent, its approach to product distribution is evolving . The company is preparing to unveil its initial AI models created under Alexandr Wang’s leadership . These models are intended to help Meta catch up to rivals after its Llama 4 models lagged significantly behind .
While Meta has established itself as the leading United States company for allowing modifications to cutting-edge AI, it is adopting a more hybrid approach moving forward . The company plans to provide versions of its upcoming models under an open-source license, aligning with Wang’s goal of democratizing access to AI . However, insiders indicate that Meta will keep certain aspects of the largest new models proprietary to retain a competitive advantage and prevent additional safety risks .
Wang believes competitors like Anthropic and OpenAI are increasingly focused on enterprise and government clients . In contrast, Meta aims to distribute its models globally to consumers, leveraging its massive existing user base across free services like WhatsApp, Facebook, and Instagram .
