Meta is scrambling to deliver its next frontier AI model, codenamed Avocado, as internal friction mounts over the company's shifting strategy from open-source Llama models to proprietary development. The social media giant's $14.3 billion bet on new AI leadership is creating cultural clashes while competitors like OpenAI and Google pull ahead in the AI race.
Meta is facing its biggest AI reckoning yet. The company's ambitious frontier model, codenamed Avocado, won't arrive until the first quarter of 2026 - a significant delay that underscores the mounting pressure on CEO Mark Zuckerberg's massive AI restructuring effort.
The delay represents more than just technical challenges. According to sources familiar with the project, Avocado is wrestling with training-related performance testing as Meta tries to ensure the system will be competitive when it debuts. "Our model training efforts are going according to plan and have had no meaningful timing changes," a Meta spokesperson told CNBC, though insiders paint a different picture of internal urgency.
The pressure stems from Zuckerberg's dramatic $14.3 billion bet on Scale AI founder Alexandr Wang, who joined as chief AI officer in June. Wang's arrival marked a seismic shift for a company that historically promoted from within. Now he's leading the elite TBD Lab, developing Avocado under intense scrutiny as OpenAI, Google, and Anthropic continue advancing their models.
Meta's AI strategy has undergone a fundamental transformation this year. The company that once championed open-source development through its Llama models is now considering a proprietary approach for Avocado. The shift came after Llama 4's disappointing April release failed to captivate developers, triggering what sources describe as a major internal shake-up.
"Today, several tech companies are developing leading closed models," Zuckerberg wrote in a blog post last July, hinting at the strategic pivot. "But open source is quickly closing the gap." His tune has changed considerably since then, particularly after Chinese AI lab DeepSeek's R1 model incorporated pieces of Llama's architecture - a development that upset many within Meta and reinforced concerns about open-source risks.
The cultural transformation has been jarring for longtime employees. Wang and former GitHub CEO Nat Friedman, who heads MSL's product arm, have brought Silicon Valley's modern AI development practices to - including 70-hour workweeks that have become the norm across AI organizations. Their "demo, don't memo" philosophy clashes with Meta's traditional collaborative approach.
