TEMPO.CO, Jakarta - Meta is developing new artificial intelligence (AI) models for images and videos, with a target release in the first half of 2026. This project is managed under Meta's superintelligence laboratory led by Alexandr Wang, a founder of Scale AI.
As reported by TechCrunch on December 19, 2025, Meta is working on image and video models under the codename 'Mango.' It is also building a text-based model internally called 'Avocado.' These plans were shared during an internal Q&A session last Thursday. The event featured Wang and Meta’s chief product officer, Chris Cox.
Wang stated that Meta wants its text-based models to have better coding skills. The company is also exploring models that can understand visual data. These models are designed to reason, plan, and act without needing training for every specific task. This approach aims to create more flexible and "human-like" AI logic.
Recently, Meta has been seen as lagging behind rivals like OpenAI, Anthropic, and Google. Meta’s AI division has faced a major reshuffle this year. This included leadership changes and hiring researchers from other big tech firms. However, the path has been rocky. Several researchers who joined the Meta Superintelligence Labs (MSL) have already left the company.
Last month, Meta's AI chief scientist Yann LeCun announced he would leave the company. He plans to start his own AI startup. His exit marks a significant shift for the company's research direction. Meta currently lacks a standout AI product that leads the market on its own merits.
So far, the growth of Meta's AI assistant relies on its massive social media base. Because the assistant is built into the search bars of popular apps, it reaches billions of people easily. Given this reality, the first projects from MSL will be vital. These models will likely decide the future of Meta's AI ambitions.
Read: Meta Reportedly Shifts AI Strategy, Develops Closed Model "Avocado"
Click here to get the latest news updates from Tempo on Google News