In a recent announcement, the CEO of Meta, Mark Zuckerberg, unveiled his vision for what he calls “personal superintelligence.” This concept revolves around the idea that artificial intelligence should empower individuals to achieve their personal aspirations and goals.
Within this communication, Zuckerberg hinted at a significant change in Meta’s approach to releasing AI models as the company embarks on its journey towards superintelligence.
He stated, “We believe that the advantages of superintelligence should be accessible to everyone globally.” However, he also acknowledged that the development of superintelligence introduces unique safety challenges. Therefore, it is crucial to be meticulous in addressing these risks and judicious in deciding which models to make open source.
This statement regarding open sourcing is particularly noteworthy. Historically, Zuckerberg has positioned Meta’s Llama series of models as a key differentiator from competitors such as OpenAI and Google DeepMind. The company’s ambition has been to create open AI models that rival or surpass those of its closed-source counterparts. In a previous letter, he expressed confidence that future Llama models would lead the industry in advancement.
However, Zuckerberg has previously indicated that he might reconsider this commitment if the capabilities of the models evolve significantly. He mentioned, “If there comes a time when we believe it is irresponsible to open source a model due to its capabilities, we will refrain from doing so.”
While some argue that Llama does not fully meet the criteria for open source AI—partly due to the absence of publicly available training datasets—Zuckerberg’s recent comments suggest a potential shift in priorities. Open sourcing may no longer be the default approach for Meta’s most advanced AI technologies.
There are compelling reasons why Meta’s competitors opt for closed models. Such models provide companies with greater control over monetization strategies. Zuckerberg noted that Meta’s business model does not rely on selling access to AI models, which means that releasing Llama does not jeopardize their revenue or research investments, unlike for companies that operate closed models. Meta primarily generates revenue through internet advertising.
However, this perspective on open models was articulated before Meta began to feel the pressure of competition and executives became increasingly focused on surpassing OpenAI’s GPT-4 model during the development of Llama 3.
Fast forward to June 2025, when Meta intensified its efforts in artificial general intelligence (AGI) by investing significantly in Scale AI, acquiring key personnel, and restructuring its AI initiatives under a new division called Meta Superintelligence Labs. The company has reportedly invested billions to attract top researchers and engineers from leading AI firms and to expand its data center capabilities.
Recent reports suggest that this influx of investment has led Meta to pause testing on its latest Llama model, Behemoth, in favor of concentrating on the development of a closed model.
With Zuckerberg’s ambition to introduce “personal superintelligence”—a clear departure from competitors focused on automating all valuable tasks—his strategy for monetizing AI is becoming clearer. It is evident from his recent statements that Meta intends to deliver personal superintelligence through its own products, such as augmented reality glasses and virtual reality headsets.
He elaborated, “Devices like glasses that can perceive our surroundings and interact with us throughout the day will become our primary computing tools.”
When questioned about the possibility of keeping their most advanced models proprietary, a Meta representative reaffirmed the company’s commitment to open source AI while also indicating plans to develop closed-source models in the future.
“Our stance on open source AI remains unchanged,” the spokesperson stated. “We will continue to release leading open source models while also training a combination of open and closed models moving forward.”
This article has been updated to include additional insights regarding Mark Zuckerberg’s position on open AI models.