Meta targets 2026 launch for new image and video AI models

Meta is fully focused on developing new AI models through its superintelligence lab, now led by Scale AI co-founder Alexandr Wang. According to The Wall Street Journal, the company is working on an image and video model called “Mango” and a new text-based model internally known as “Avocado.”

The tech giant plans to launch these models in the first half of 2026. The roadmap was revealed during an internal Q&A at Meta on Thursday by Wang and Chief Product Officer Chris Cox.

Wang stated that Meta aims to improve the text-based model’s coding abilities while exploring new “world models” that can understand visual information, reason, plan, and act without needing training on every possible scenario.

The push comes as Meta attempts to catch up with competitors like OpenAI, Anthropic, and Google.

The company’s AI division underwent significant restructuring this year, including leadership changes and aggressive recruitment of researchers from rival firms. However, retention has been an issue, with several researchers who joined Meta Superintelligence Labs (MSL) having already left.

In a major blow to the division, the company’s chief AI scientist, Yann LeCun, revealed in November that he is leaving to create his own startup.

While usage figures for Meta’s AI assistants are high—boosted by integration into the search bars of Facebook, Instagram, and WhatsApp—the company has yet to launch a truly successful, standalone AI product.

This raises the stakes significantly for the first projects emerging from MSL, as Meta looks to prove it can lead the next wave of AI innovation.