Meta is developing a new generation of artificial intelligence models for release in the first half of 2026, according to a report by The Wall Street Journal. The effort is being led by the company’s superintelligence lab, headed by Scale AI co-founder Alexandr Wang.
The company is working on an image and video model internally known as Mango, alongside a text-based model codenamed Avocado. The roadmap was outlined during an internal company session featuring Wang and Meta chief product officer Chris Cox. Wang said the text model is being designed with stronger coding capabilities, while the visual model will focus on understanding images and video in ways that enable reasoning, planning, and action.
Meta’s renewed push follows a period of disruption within its AI division, including leadership changes and researcher turnover. The company has lagged behind rivals such as OpenAI, Google, and Anthropic in releasing widely adopted AI models. Its Meta AI assistant currently relies heavily on distribution through Meta’s social platforms rather than standalone performance advantages.
The upcoming models are expected to play a central role in Meta’s AI strategy. With no breakout AI product yet, early releases from the superintelligence lab carry heightened expectations as Meta seeks to regain momentum in the competitive AI landscape.
