Meta is quietly working on a major expansion of its artificial intelligence efforts. Reports suggest that the company is developing three new AI models at the same time, covering images and videos, software coding, and advanced world simulation. These projects are being handled by Meta’s Superintelligence Labs, a dedicated AI division the company set up earlier this year to compete more aggressively with rivals like Google and OpenAI.
New AI models for images and videos
According to a report by The Wall Street Journal, one of Meta’s biggest upcoming projects is an image and video generation AI model internally called “Mango.” Meta’s Chief AI Officer and head of Superintelligence Labs, Alexandr Wang, reportedly shared details about this model during an internal company discussion.
Mango is expected to launch in the first half of 2026 and will focus on creating high quality images and videos. The model is designed to compete directly with tools such as Google’s Nano Banana and Veo, along with OpenAI’s image generation and Sora video models.
A coding focused AI model called Avocado
Alongside Mango, Meta is also building a new text based AI model focused entirely on coding. This project is internally known as “Avocado” and is expected to follow a similar release timeline in 2026.
With Avocado, Meta is entering the fast growing market of AI coding assistants. Other tech giants have already launched similar tools, including Google, OpenAI, and Anthropic. Industry experts believe Meta could eventually use this model to build its own developer tools, such as a coding assistant, plugin, or even a full integrated development environment.
Big progress in world simulation AI
The report also points to major progress in a more advanced area of AI called world models. Meta has been researching this technology for years, especially under former AI chief Yann LeCun. World models allow AI systems to learn from visual data and simulate how the real world works, including physics and cause and effect.
These models help AI predict what might happen before taking action, which is seen as a key step toward more intelligent systems. This technology is especially important for the future of robotics, where machines need to understand and interact with real world environments safely and accurately.
Overall, Meta’s push into visual AI, coding assistance, and world simulation shows that the company is aiming far beyond chatbots. If these projects succeed, they could play a major role in Meta’s long term goal of building more general and capable artificial intelligence systems.
