TEMPO.CO, Jakarta - Tech giant Meta is reportedly developing a new artificial intelligence (AI) model dubbed "Avocado," which could be a significant departure from the company's open-source strategy. This move comes as Mark Zuckerberg has been distancing himself from the open model approach as the company focuses on developing superintelligence.
According to Engadget, Avocado might turn out to be a proprietary model instead of open source. This model is being developed by a small unit named TBD under Meta's AI Superintelligence Labs, led by Chief AI Officer Alexandr Wang, and is scheduled for release in 2026.
It is not clear what this project means for the future of Llama. Zuckerberg previously stated that Meta would continue to lead in open source, but the company will not "open source everything that we do." He also expressed concerns about security related to superintelligence.
This shift in strategy is also triggered by setbacks in Llama 4. The Llama 4 Behemoth model has been delayed for months, and The New York Times reported that Wang and several executives have discussed canceling it altogether. Internal developers are also said to be dissatisfied with the available version of Llama 4.
Meta is also making a major overhaul in the company's AI team structure. The company has laid off several hundred employees in the Fundamental Artificial Intelligence Research (FAIR) unit. Yann LeCun, Meta's Chief AI Scientist, known as a strong advocate of open source, recently announced that he is leaving the company.
The shift towards closed models is a significant move for Zuckerberg. Last year, he said "fuck that" about closed platforms and published a memo entitled "Open Source AI is the Path Forward." However, he is said to be intensely worried that Meta is falling behind OpenAI, Google, and other competitors. The company is estimated to spend US$600 billion in the coming years to fund its AI ambitions.
Read: OpenAI's Pocket AI Device Could Redefine Computing