Mark Zuckerberg has had many months publicly initiated that he is backing from open source AI. Now, Meta’s latest AI Pivot is starting to come into focus. The company is reportedly working on a new model, known within Meta as “avocado,” which could mark a major shift in AI development.

BOTH CNBC and Bloomberg reported on Meta’s plans surrounding “avocado,” with both outlets saying the model “could be” proprietary rather than open-source. Avocado, due now in 2026, works within “TBD,” a small group within the AI ​​Superintelligence Labs headed by Chief Alexandr Wang, who apparently prefers closed models.

It is not clear what Avocado means for Llama. Last year, Zuckerberg said he expected Meta to “continue to be a leader” in open source but that it would not “open source everything we do.” He also cited safety concerns as they relate to superintelligence. Like both CNBC and Bloomberg Remember, Meta’s move was also driven by issues surrounding the release of LLAMA 4. The Llama 4 “Behemoth” model delayed for many months; The New York Times reported Last year Wang and other execs had “discussed abandoning” it all. And the developers are reportedly unrestricted by the Llama 4 models available.

There have been other shake-ups within the ranks of AI groups with Zuckerberg spent billions of dollars building a dedicated team of superintelligence. The company set up several hundred workers from the basic artificial intelligence research (fair) unit. And meta veteran and chief scientist Yann Lecun, who is a proponent for open source and skeptical of Llms, recently informed him LEAVE the company.

That meta could pursue a closed AI model is a significant shift for Zuckerberg, who last year said that “Fuck that“About closed platforms and wrote a long memo titled” Open Source AI is the Way of the Soul, Google and other opponents. Meta says it will cost money $ 600 billion Over the next few years to fund AI ambitions.