Alexandr Wang, a co-founder of Scale AI, is currently in charge of Meta’s superintelligence lab, which is entirely focused on creating new AI models. The Wall Street Journal reports that the corporation is developing an internal text-based model named “Avocado” and an image and video model called “Mango.” In the first half of 2026, the tech behemoth intends to introduce these models. Wang and Chief Product Officer Chris Cox unveiled the roadmap on Thursday during an internal Q&A at Meta.
According to Wang, Meta seeks to enhance the text-based model’s coding capabilities while investigating novel “world models” that are capable of comprehending visual data, reasoning, planning, and acting without requiring training on every scenario. Meta is making this drive in an effort to overtake rivals like Google, Anthropic, and OpenAI. This year, the company’s AI division underwent a major restructuring that included aggressive recruiting of experts from competing companies and managerial changes. Retention has been a problem, though, as many of the researchers that joined Meta Superintelligence Labs (MSL) have already departed. The division suffered a significant setback when Yann LeCun, the company’s chief AI scientist, announced in November that he was quitting to found his own firm.
Although Meta’s AI assistants have a high utilization rate due to their inclusion into Facebook, Instagram, and WhatsApp search bars, the business has not yet introduced a really successful stand-alone AI product. As Meta seeks to demonstrate that it can spearhead the next generation of AI innovation, this increases the stakes considerably for the initial initiatives coming out of MSL.
