The company is developing new AI models, including an image and video model called “Mango” and a text model internally called “Avocado.”
The Meta Platform logo seen in Davos, Switzerland, May 22, 2022. Photographed on May 22, 2022. Reuters/Arnd Wiegmann
“>
The Meta Platform logo seen in Davos, Switzerland, May 22, 2022. Photographed on May 22, 2022. Reuters/Arnd Wiegmann
In the era of AI competition, Meta has struggled to keep pace, and its next move is targeting AI-generated video, which is rapidly growing in the competition.
Meta plans to release a new AI model focused on images and video alongside the next version of its text-based model, The Wall Street Journal reported.
The company is developing new AI models, including an image and video model called “Mango” and a text model internally called “Avocado.”
The project is led by Meta's chief AI officer, Alexandr Wang, founder of Scale AI, who joined Meta earlier this year.
The social media company plans to release a new model in the first half of 2026. Wang, along with Chris Cox, Meta's chief product officer, announced the new roadmap.
Meta will face stiff competition in an already crowded and increasingly saturated field of creative tools for generating images and videos. Alphabet's Google took an early lead with products like Veo and Nano Banana, followed by OpenAI's Sora and Adobe's Firefly. But Meta enters this race with an important advantage. It's a large user base deeply embedded across social platforms that could help drive rapid adoption despite a crowded market.
Wang said Meta is focused on enhancing its text-based models and improving its coding capabilities, while also working on developing new world models that can understand visual information and be able to reason, plan and act on all possible scenarios without training.
The company is planning a major reorganization of its AI division this year, including a change in leadership and aggressive hiring of top researchers from rival companies. Meta CEO Mark Zuckerberg has begun a hiring spree to create a new division, Meta Superintelligence Lab (MSL), tasked with developing next-generation AI models. The division will be led by Wang, the former founder of data labeling startup Scale AI, which Meta acquired a 49% stake in for $14 billion.
However, several of the researchers who joined MSL have already left. Last month, Yann LeCun, the company's chief AI scientist, also announced he was leaving to start his own startup.
Meta doesn't have any notable AI products yet. Rather, the Meta AI assistant's popularity comes primarily from Meta's huge social media user base, which includes billions of people. By putting Assistant directly in the app's search bar, the company is making sure users see and use Assistant more because of its reach and placement, rather than because the product itself is clearly better.
This means that MSL's initial projects and models bring great expectations and pressure in the AI race.
