Meta’s AI Ambitions Pivot Amid Llama Setbacks – What Is The Proprietary Avocado Model?

• CEO Mark Zuckerberg initially projected that the company’s Llama AI models would set the standard for the industry and democratize AI access. 
• Meta executives are addressing Avacado’s performance testing and model training challenges.
• In 2025, Meta undertook a multibillion-dollar hiring spree, bringing in top AI talent.

Meta Platforms Inc. (META) is reportedly reshaping its AI strategy after recent developments indicate a shift from open-source Llama initiatives toward a more proprietary and high-investment approach, aimed at competing directly with OpenAI, Google, and Anthropic.

According to a CNBC report, CEO Mark Zuckerberg initially projected that the company’s Llama AI models would set the standard for the industry and democratize AI access. 

Project Avacado 

Meta is now working on a next-generation AI project, codenamed Avocado, which is expected to launch in early 2026, rather than the initially anticipated late-2025 timeline, said the report. 

Meta executives are addressing performance testing and model training challenges to ensure a smooth rollout, signaling a more cautious approach than last year’s ambitious Llama announcements.

Meta’s stock inched 0.4% lower in Tuesday’s premarket. On Stocktwits, retail sentiment around the stock remained in ‘neutral’ territory amid ‘normal’ message volume levels. 

Meta’s early advantage in open-source AI is waning, according to the report. After Llama 4’s lukewarm reception and unauthorized use of its architecture by competitors, the company is reconsidering open-source strategies. Avocado is expected to be a proprietary system that restricts external access to its models, reinforcing Meta’s push for security and control over its AI technology.

Leadership Shake-Up And Talent Acquisition

In 2025, Meta undertook a multibillion-dollar hiring spree, bringing in top AI talent, including Alexandr Wang, founder of Scale AI, and other industry veterans like Nat Friedman and Shengjia Zhao. 

Wang now leads the new TBD Lab, which manages Meta’s large language models. These hires mark a departure from Meta’s historic preference for long-tenured internal promotions.

The company has also planned substantial budget reductions for its metaverse division, signalling a strategic shift in vision towards AI and related tools. 

META stock has gained over 12% in 2025 and over 6% in the last 12 months. 

Also See: Satya Nadella Announces Microsoft’s $17.5 B Commitment To Build India’s AI Infrastructure

For updates and corrections, email newsroom[at]stocktwits[dot]com.

Share

·

Add us on