Zuckerberg: Meta Won't Open Source All AI Models

Meta's Evolving AI Strategy: A Shift Towards "Personal Superintelligence"

Mark Zuckerberg, CEO of Meta, outlined his vision for “personal superintelligence” on Wednesday, describing a future where artificial intelligence empowers individuals to achieve their specific objectives.

A Change in Approach to AI Model Release

Embedded within Zuckerberg’s statement is an indication of a strategic adjustment in Meta’s plans for releasing AI models as the company progresses in its pursuit of “superintelligence.”

Zuckerberg emphasized the importance of broadly sharing the benefits of superintelligence. However, he also acknowledged the emergence of new safety concerns that will necessitate careful risk mitigation and a cautious approach to open-sourcing.

The Significance of Open Source

This nuanced perspective on open source is noteworthy. Historically, Meta has differentiated itself from competitors such as OpenAI, xAI, and Google DeepMind through its Llama family of open models.

Meta’s ambition has been to develop open AI models that are comparable to, or even surpass, the capabilities of closed models. Zuckerberg previously predicted that future Llama models would become the most advanced in the industry starting in 2024.

Maintaining Flexibility

Zuckerberg has consistently maintained a degree of flexibility regarding this commitment. He stated that if a qualitative shift in a model’s capabilities raises concerns about responsible open-sourcing, Meta would refrain from doing so.

Despite Meta’s efforts, some argue that Llama doesn’t fully meet the criteria for truly open-source AI, particularly due to the lack of public access to its extensive training datasets.

A Potential Shift in Priorities

Zuckerberg’s recent statements suggest a possible change in priorities, indicating that open source may no longer be the automatic default for Meta’s most advanced AI developments.

The Advantages of Closed Models

Meta’s competitors often favor closed models because they provide greater control over the monetization of their products.

Zuckerberg previously highlighted that Meta’s business model isn’t dependent on selling access to AI models, meaning that releasing Llama doesn’t negatively impact its revenue or research investment, unlike closed-source providers.

Meta primarily generates revenue through internet advertising.

The Impact of Competitive Pressure

This viewpoint on open models existed before Meta began to perceive itself as lagging behind competitors, and executives became focused on surpassing OpenAI’s GPT-4 model with Llama 3.

Investment and Restructuring

In June 2025, Meta initiated a significant push towards achieving artificial general intelligence (AGI) by investing $14.3 billion in Scale AI, acquiring its founder and CEO, and establishing a new division called Meta Superintelligence Labs.

The company has allocated billions of dollars to recruit researchers and engineers from leading AI firms and construct new data centers.

Focus on Closed Model Development

Recent reports indicate that Meta has temporarily halted testing of its latest Llama model, known as Behemoth, to concentrate its resources on developing a closed model.

Monetization Strategy and "Personal Superintelligence"

With Zuckerberg’s focus on introducing “personal superintelligence” – a departure from competitors aiming to “automate all valuable work” – his AI monetization strategy is becoming clearer.

Zuckerberg’s recent communication confirms Meta’s intention to deliver “personal superintelligence” through its own products, such as augmented reality glasses and virtual reality headsets.

“Personal devices like glasses that understand our context because they can see what we see, hear what we hear, and interact with us throughout the day will become our primary computing devices,” Zuckerberg explained.

Meta's Stance on Open Source Remains

When questioned about the possibility of keeping its most advanced models closed, a Meta spokesperson reiterated the company’s commitment to open-source AI.

The spokesperson also stated that Meta anticipates continuing to train both open and closed-source models in the future.

“Our position on open source AI is unchanged,” the spokesperson affirmed. “We plan to continue releasing leading open source models. We haven’t released everything we’ve developed historically and we expect to continue training a mix of open and closed models going forward.”

This article has been updated to reflect further details regarding Mark Zuckerberg’s position on open AI models.