When a company starts assigning fruits as codenames for AI models, it is an indicator that two things are happening simultaneously, the laboratory workers are having a lot of fun, and very high stakes are involved.

To the casual observer, Meta’s latest AI initiatives, named  “Mango” and “Avocado” may seem to be silly, but under the surface of fun names is a genuine effort to reclaim the lost ground in an AI race that the company never intended to be part of.

Meta’s Superintelligence Lab

Meta is softly preparing for an AI major update, where it is working on better image, video, and text models for early 2026 release, as per The Wall Street Journal.

This is the initiative of Meta’s Superintelligence Lab, which is led by Alexandr Wang, the co-founder of Scale AI, who is now guiding the company’s most ambitious AI plan so far.

Moreover, Meta is simultaneously developing an image and video model that is codenamed as “Mango”, along with a text-focused model that is referred to as “Avocado.” This plan was reportedly disclosed during an internal Q&A session in which Wang and Meta’s chief product officer, Chris Cox, illustrated how these models are to shape the company’s upcoming AI strategy.

Beyond Chatbots

The goal is not just all about the chatbots, but also in “world models”.Wang says that the new text-based model of Meta is meant to bring about a remarkable enhancement in coding skills, which is a major aspect where the likes of OpenAI and Anthropic are ahead.

At the same time, the tech giant is progressing further through the development of “world models,” which can not only process the visual data and reason about it, but can also plan and act without being explicitly trained in every possible circumstance.

This move indicates the transition from simple assistants to more independent multimodal systems, an AI that not only reacts but also interprets and predicts. If they are able to pull it off, then Mango and Avocado might actually be Meta’s first genuine step into the world of general-purpose intelligence.

Catching up with AI

The company’s AI division had a wild time, and as a result, Meta refocused its efforts. They have moved to the back of the line behind the likes of OpenAI, Google, and Anthropic, which led them to restructure, change management, and hire aggressively from other labs. Still, the transitions have not been smooth.

A number of top researchers who were recruited to the SuperIntelligence Lab at Meta have left, and their resignations have been a source of instability within the organization. Last month, the uncertainty of the situation worsened when Yann LeCun, the chief AI scientist at Meta, decided to leave the company to start his own venture.

Although Meta’s technology and staff are among the best, the problem of instability has been there all the time.

Not a Single AI Breakthrough So Far

Though Meta has a huge audience, there is no single AI product from the company that can be said to lead the market explicitly. The number of users of Meta AI mainly comes from its distribution rather than the demand for it, because the assistant is integrated into the search bars of Facebook, Instagram, and WhatsApp.

The magnitude of the user base is massive, but that does not necessarily mean that the company is the technology leader. This situation makes it even more difficult for the first models released by the SuperIntelligence Lab. Mango and Avocado will not only be new products, but also the company’s credibility that is going to be tested.

Bottom Line

Mango and Avocado will likely be major reductions or merely the company’s AI rebuilding phase, as per Meta’s AI roadmap for 2026. This suggests a firm that is aware of its lost ground, but is still quite certain of its ability to transform the story.

The firm with new management, ambitious technical objectives, and billions of users as an inbuilt audience has a very large upside. Whether Mango and Avocado become breakthroughs or just notes will likely be the deciding factor of whether or not Meta regains its AI momentum.