Meta's New AI Venture: 'Avocado' Could Shift Away from Open Source

Mark Zuckerberg's New Direction in AI Development

For the past several months, Mark Zuckerberg has hinted at a departure from distributing AI models as open source. The latest direction for Meta’s artificial intelligence ambitions is becoming clearer as the company is reportedly invested in developing an AI model internally referred to as ‘Avocado’. This project may signify a substantial transformation from Meta's historical preference for open-source AI.

Reports from renowned media outlets like CNBC and Bloomberg suggest that 'Avocado' could potentially be a proprietary model, rather than one accessible to the public as open-source. Expected to launch in 2026, the project is being cultivated by a specialized team named ‘TBD’ within Meta’s AI Superintelligence Labs, led by Chief AI Officer Alexandr Wang, known for favoring closed model frameworks.

Implications and Concerns for the Llama Series

The unveiling of Avocado raises questions about its impact on Meta’s existing Llama models. Earlier, Zuckerberg expressed confidence in Meta’s continued leadership in open-source arenas, yet acknowledged not all advances would remain open to the public, citing potential risks tied to superintelligence. Compounding this is the delayed release of Llama 4—dubbed the 'Behemoth'—which has faced setbacks and internal debates over its viability, with some leaders considering its cancellation. Feedback from developers so far has not been optimistic, adding another layer of intrigue to these developments.

Restructuring and Strategic Realignments at Meta

Significant personnel changes at Meta highlight the reorganization in the company’s pursuit of superintelligence. The company recently executed layoffs affecting hundreds within its Fundamental Artificial Intelligence Research (FAIR) department. Furthermore, esteemed AI scientist Yann LeCun, a strong proponent for open-source methodologies and a critic of large language models (LLMs), has announced his departure from Meta.

This strategic pivot towards potentially non-open-source AI represents a noteworthy reversal for Zuckerberg. Not long ago, he adamantly opposed closed systems, emphasizing open-source solutions as a strategic pathway. Yet, facing fierce competition from industry titans like OpenAI and Google, Meta recognizes the necessity of advancing its AI offerings to remain competitive. The company forecasts spending upwards of $600 billion over the forthcoming years to support its ambitious AI goals.