At Meta, everything is hands-on as we develop new AI models under the Superintelligence Lab led by Scale AI co-founder Alexandr Wang. The company is currently working on an image and video model codenamed “Mango,” and a new text-based model internally called “Avocado,” The Wall Street Journal reported.
The tech giant plans to launch new models in the first half of 2026, the paper said, citing an internal Q&A with Meta on Thursday, in which Wang and chief product officer Chris Cox announced a new roadmap.
Wang said Meta aims to improve the coding capabilities of text-based models, while at the same time exploring new world models that can understand visual information and be able to reason, plan and act on all possibilities without being trained.
Meta has recently fallen behind rivals like OpenAI, Anthropic, and Google in the AI race. The company’s AI division underwent a major reorganization this year, including leadership changes and poaching of researchers from other top companies. However, several of the researchers who joined Meta SuperIntelligence Labs (MSL) have already left.
Last month, the company’s chief AI scientist, Yan Lekun, also announced he was leaving to found his own startup.
Meta currently doesn’t have a winning AI product. Instead, the Meta AI assistant’s numbers are powered by the company’s existing social network, which spans billions of users, as the company places the assistant in the app’s search bar.
This means that the initial projects and models coming out of MSL will be affected in a number of ways.
