Meta Shifts AI Strategy: Prioritizing Closed Models Under Alexandr Wang

Meta is reportedly poised for a significant Meta AI strategy change. The company appears to be moving away from an exclusively open-source model for its frontier artificial intelligence, at least initially. This strategic reversal under Alexandr Wang’s leadership has fueled growing speculation.

What Happened

Meta is preparing to release new closed source AI models. Alexandr Wang, CEO and founder of Scale AI, developed these. The company may initially offer few versions with an open-source license.

This approach marks a potential departure from Meta’s prior stance. Meta was the largest U.S. company allowing modification of its most advanced general-purpose models. Meta will initially release closed models to manage safety risks and protect specifications.

Details From Sources

Alexandr Wang’s Role and Scale AI Investment

Alexandr Wang drives key AI decisions at Meta, mediapost.com reported. Meta invested over $14 billion in Scale AI in June 2025. This acquired a 49% non-voting stake, valuing Scale AI at $29 billion.

Wang joined Meta to lead a new team focused on artificial general intelligence. This AI understands or learns any human intellectual task. He founded Scale at 19 after dropping out of MIT, now serving as Meta’s chief AI officer. His vision centers on personal superintelligence. This AI understands a user’s specific goals. This aligns with collecting data for targeted objectives, not general data gathering.

“Avocado” Model Development and Performance

Mediapost.com reported that The New York Times stated delays concerned Meta’s new foundational AI model. It is code-named “Avocado.” This model was originally slated for release in March 2026.

“Avocado” reportedly did not perform as well as rival offerings from Google and other companies. This performance issue was noted in coding, reasoning, and writing. The New York Times suggested Meta might temporarily license Gemini to power its AI products.

“Avocado” signifies a shift from the open-source Llama series towards a more powerful, potentially proprietary “frontier” model. Aspirations for open-sourcing the model still remain.

Meta’s Competitive Stance and Consumer Focus

Meta’s models developed under Wang will prioritize consumer-facing features. This includes advanced shopping tools, mediapost.com reported. These new models aim to help Meta close its performance gap with industry rivals.

According to mediapost.com, Axios reported that Meta’s first family of models is designed to help it catch up. Its previous Llama 4 family significantly underperformed. Meta believes its new models will offer distinct strengths appealing to consumers.

Why This Matters

This potential Meta AI strategy change could signify a notable shift in the broader artificial intelligence landscape. This is especially true for developers and companies relying on Meta’s open-source frontier models. It indicates a strategic move by Meta to enhance competitive performance and address safety concerns with advanced AI models.

Background Context

Meta has historically been the largest U.S. company to allow others to modify its frontier models. These are the most advanced general-purpose artificial intelligence models. The company’s Llama series gained recognition for its broad accessibility and open-source nature.

Related Data or Statistics

Meta’s investment in Scale AI amounted to more than $14 billion for a 49% non-voting stake. At the time of this investment, Scale AI was valued at $29 billion.

Future Implications (SPECULATIVE)

The shift towards initial closed source AI models could influence future AI model licensing agreements across the industry. This might also impact data collection strategies within the AI sector.

Wang’s vision of Meta superintelligence, focused on personal goals, could shape Meta’s future AI product development. Meta’s prioritization of consumer-facing features may lead to new applications, such as advanced shopping tools.

Conclusion

Meta’s evolving Meta AI strategy change under Alexandr Wang marks a pivotal moment for the company. It balances potential open-source aspirations with initial closed model releases. The strategy emphasizes competitive performance and enhanced consumer-facing features in the AI market.

FAQ Section

Q1: What is Meta’s rumored new approach to AI models?

A1: Meta is reportedly shifting towards initially releasing closed AI models. This marks a potential change from its previous open-source strategy for frontier models.

Q2: Who is Alexandr Wang and what is his role at Meta?

A2: Alexandr Wang is the CEO and founder of Scale AI. He drives AI decisions at Meta. He leads a new team focused on artificial general intelligence. He serves as Meta’s chief AI officer.

Q3: What is “Avocado” and what were its reported issues?

A3: “Avocado” is the code-name for Meta’s new foundational AI model. Mediapost.com reported that The New York Times stated it experienced delays. It reportedly did not perform as well as rivals in coding, reasoning, and writing.

Q4: How much did Meta invest in Scale AI?

A4: Meta invested over $14 billion in Scale AI in June 2025. This acquired a 49% non-voting stake, which valued Scale AI at $29 billion.

Q5: What is Alexandr Wang’s vision for “personal superintelligence”?

A5: Wang’s vision for personal superintelligence involves AI that understands a user’s specific goals and interests. This aligns with collecting data for targeted purposes like event planning or ad campaigns.