Meta’s Secret AI Models Mango and Avocado Target 2026 Release for Industry Lead

Meta is developing powerful new AI models under a secretive superintelligence lab. The projects, codenamed “Mango” and “Avocado,” aim for release in early 2026. This was revealed in an internal company meeting this week.

The tech giant confirmed its roadmap to staff. According to The Wall Street Journal, Scale AI co-founder Alexandr Wang and Chief Product Officer Chris Cox led the presentation. Their goal is to catch up to rivals like OpenAI and Google.

Inside the Development of Next-Generation AI

The “Mango” model focuses on images and video. “Avocado” is a text-based model with enhanced coding capabilities. Both projects are core initiatives of Meta’s Superintelligence Labs (MSL).

Wang stated the aim is to create AI that can reason and plan. The models would understand visual information without exhaustive training. This represents a significant leap beyond current assistant technology.

Meta’s Uphill Battle in the Competitive AI Race

Meta has fallen behind key competitors recently. The company has restructured its AI division multiple times this year. It also recruited top researchers from other firms to bolster its efforts.

Several of those new hires have already departed MSL. Last month, Chief AI Scientist Yann LeCun also announced his exit to start his own company. These shifts highlight internal challenges.

Currently, Meta lacks a standalone winning AI product. Its Meta AI assistant is distributed through its social apps like Facebook and Instagram. This provides user reach but not necessarily technological leadership.

The success of Mango and Avocado is now critical. They carry the weight of Meta’s ambition to lead the industry. Their 2026 debut will be a major test for the company’s revised strategy.

The development of Mango and Avocado marks a pivotal investment for Meta. The company’s future position in the AI industry may hinge on these 2026 models. All eyes are now on its superintelligence lab to deliver.

A quick knowledge drop for you

What are Meta’s new AI models called?

Meta is developing two new models. They are internally codenamed “Mango” for image and video, and “Avocado” for advanced text and coding tasks.

When will these AI models be released?

The target release window is the first half of 2026. This was announced internally by Meta executives to their team this week.

Why is Meta creating these new models?

Meta aims to close the gap with leaders like OpenAI and Google. The company seeks to build AI that can reason and plan more independently.

Who is leading this development at Meta?

The effort is run by Meta’s Superintelligence Labs. It is led by Alexandr Wang, the co-founder of Scale AI, alongside top Meta executive Chris Cox.

Does Meta currently have a leading AI product?

Not yet. Its Meta AI assistant is accessible but is not considered a market leader. Its user numbers are boosted by integration into Meta’s popular social apps.

Has there been recent staff changes in Meta’s AI lab?

Yes. The division has seen restructuring and some high-profile researcher departures. Chief AI Scientist Yann LeCun recently left to start his own company.