In what could represent a radical shift in Meta’s strategy, members of its newly formed superintelligence lab, including new Head of AI Alexandr Wang, have discussed abandoning the company’s most powerful open source AI model in favor of developing a closed model.
Introduction: A potential strategic shake-up
Meta’s newly formed superintelligence lab has discussed making a series of changes to the company’s artificial intelligence strategy, in what could amount to a major shake-up for the social media giant. Last week, a small group of senior lab members, including 28-year-old Alexandr Wang, Meta’s new Head of AI, discussed abandoning the company’s most powerful open source AI model, named Behemoth, in favor of developing a closed model, according to two people familiar with the matter.
A philosophical and technical shift
For years, Meta has chosen to open source its artificial intelligence models, meaning it makes the computer code public for other developers to build upon. Closed models keep their underlying code proprietary. Meta executives have long argued that it is better to build technology openly so that AI development moves faster and is available to more developers.
Any move toward a closed AI model would be as much a philosophical change at Meta as it is a technical one. Meta has won praise from developers for open-sourcing its AI models, and one of its top AI executives, Yann LeCun, has said that “the platform that will win will be open.” This year, Chinese artificial intelligence company DeepSeek released an advanced chatbot thanks in part to Meta’s open-source code.
The mysterious fate of the Behemoth model
Meta has finished data entry to fine-tune its Behemoth model—a process known as “training”—but has delayed its release due to poor internal performance, according to people familiar with the matter. After the company announced the formation of the superintelligence lab last month, teams working on the Behemoth model—known as the “frontier” model—stopped running new tests on it, one person said.
Discussions by the superintelligence lab are preliminary, and no decisions have been made regarding potential changes, which would require the approval of Meta CEO Mark Zuckerberg. Meta could maintain open-source AI models while prioritizing a closed model. If these scenarios occur, they would mark a major pivot for the company as it attempts to remain competitive in the AI race against rivals such as Google, OpenAI, and Anthropic.
Superintelligence Lab: An elite team
Meta’s superintelligence lab is being closely watched after the company recently stumbled in AI technology. Zuckerberg has embarked on a spending spree to create the new lab, offering nine-figure compensation packages to hire top researchers. In June, the company invested $14.3 billion in AI startup Scale AI, founded and led by Wang. Under the deal, Meta acquired a 49 percent stake in the company, and Wang and a team of top Scale employees joined Meta in leadership roles.
The company has since renamed its entire AI division “Meta Superintelligence Labs,” with Wang serving as Head of AI. Within the larger AI division, Wang has led an exclusive team of about a dozen newly hired researchers, a few of his deputies from Scale AI, and Nat Friedman, former CEO of GitHub.
Conclusion: An uncertain future
These internal discussions represent a critical moment for Meta. The decision to abandon open source would not just be a technical change, but a retreat from a core philosophy the company has championed for years. Clearly, the pressure to compete at the highest level in the AI race is pushing Zuckerberg and his new team to re-evaluate everything. The decisions made in the coming months will determine the trajectory of Meta’s AI efforts for years to come and could permanently alter its relationship with the broader developer community.
SEO Title:
Meta Description:
