Mark Zuckerberg has been publicly hinting for months that he’s moving away from open-source AI models. Now, Meta’s new, key direction in AI is starting to become clear. The company is developing a new model, internally called Avocado. This could be a major departure from the company’s previous strategy of developing AI on the basis of open source.
Mark Zuckerberg has been publicly hinting for months that he’s moving away from open-source AI models. Now, Meta’s new, key direction in AI is starting to become clear. The company is developing a new model, internally called Avocado. This could be a major departure from the company’s previous strategy of developing AI on the basis of open source.
The new Avocado model is expected to debut next spring and could launch as a closed model — one that can be tightly controlled and to which Meta can sell access, Bloomberg reports , citing its own sources.
Avocado is being developed by an internal TBD group, a small division within AI Superintelligence Labs. The group is led by Chief Artificial Intelligence Officer Alexander Wang , who is a proponent of closed models.
As Bloomberg notes, Meta's change in strategy was also prompted by issues surrounding the release of Llama 4. The release of the Llama 4 Behemoth model has been delayed for months. The New York Times reported earlier this year that Wang and other executives were "discussing scrapping it altogether." Additionally, developers were reportedly not thrilled with those Llama 4 models.
That Meta is now potentially moving to a closed AI model is a significant shift for Zuckerberg, who only last year spoke out strongly against closed platforms and wrote a lengthy memo titled “Open Source AI is the Way Forward.” But he is clearly deeply concerned about falling behind rivals like OpenAI and Google. Meta has said it plans to invest $600 billion over the next few years to pursue its AI ambitions.
