For several months, Mark Zuckerberg has openly suggested a retreat from fully open-source AI initiatives. Details are emerging about Meta's evolving strategy in artificial intelligence, including the development of an internal project codenamed 'Avocado,' signaling a potential departure from the firm's longstanding commitment to sharing AI technologies publicly.
Sources including CNBC and Bloomberg have detailed Meta's intentions for Avocado, noting that the upcoming system, expected to launch around 2026, might remain closed-source rather than freely available. This effort is being spearheaded within a specialized unit called 'TBD,' part of Meta's AI Superintelligence Labs, under the leadership of Chief AI Officer Alexandr Wang, who is known to support proprietary approaches to AI.
The future of Meta's Llama series in light of Avocado remains uncertain. Earlier this year, Zuckerberg affirmed that Meta intended to maintain its prominence in open-source AI while clarifying that not all projects would be released openly. He has also expressed apprehensions about safety risks associated with advanced AI capabilities. According to the same reports, Meta's strategic change has been influenced by challenges with Llama 4, whose massive 'Behemoth' variant has faced prolonged delays. The New York Times revealed that Wang and senior leaders considered scrapping it entirely, and feedback from developers on existing Llama 4 versions has been underwhelming.
Amid these developments, Meta's AI divisions have undergone notable restructuring as Zuckerberg invests heavily in superintelligence pursuits. The firm recently cut several hundred positions in its Fundamental Artificial Intelligence Research (FAIR) team. Additionally, longtime Meta executive and Chief AI Scientist Yann LeCun, a vocal advocate for open-source methods and a critic of large language models, has decided to depart the organization.
This move toward a non-open AI project represents a stark change for Zuckerberg, who only a year ago dismissed proprietary systems emphatically and authored an extensive document advocating 'Open Source AI is the Path Forward.' Nevertheless, the highly ambitious leader appears deeply concerned about lagging behind competitors like OpenAI and Google. Meta has projected expenditures of up to $600 billion in the coming years to advance its AI objectives.