Meta CEO Mark Zuckerberg’s ambitious plans for the Llama AI models have taken a backseat. Once hailed as “the most advanced in the industry,” Llama is now mentioned only occasionally, as Meta focuses on building a stronger, proprietary AI lineup.
New AI Project: Avocado
Meta is developing a new AI model, codenamed Avocado, expected to launch in early 2026. The model is undergoing extensive training and testing to ensure it performs well.
Meta has recruited top AI talent, including Scale AI founder Alexandr Wang, to lead the project. The company spent $14.3 billion on these hires and raised its 2025 capital expenditure guidance to $70–72 billion.
Moving Away from Open Source
Llama was originally open source, allowing external developers to access and improve the technology. Setbacks like the poor reception of Llama 4 and concerns over rivals using its architecture have prompted Meta to rethink this approach. Avocado is expected to be proprietary, keeping its software closed to outside developers.
Competing in a Fast-Moving Market
Meta faces growing pressure from rivals. Google’s Gemini 3, OpenAI’s GPT-5 updates, and Anthropic’s Claude Opus 4.5 are advancing quickly.
Internal teams work long hours under tight deadlines, while restructuring and layoffs have shaken the AI division.
Zuckerberg has shifted leadership to outsiders like Wang and former GitHub CEO Nat Friedman, focusing on infrastructure and frontier AI development.
The company is also prioritizing AI-driven products over virtual reality projects and adopting faster development methods, emphasizing demonstrations over long internal approvals.
Meta’s AI pivot is a high-stakes gamble. With billions invested and top talent on board, the company aims to regain momentum and compete with leading AI developers, while continuing to strengthen its advertising business.