When companies start assigning fruit as code names to their AI models, it shows that two things are happening at the same time: lab employees are having a lot of fun, and there are very high stakes involved.
For the casual observer, the meta Latest AI initiativesThe names “Mango” and “Avocado” may seem silly, but beneath the surface of the fun names lies a real effort to regain lost ground in the AI race the company never intended to be a part of.
Meta's Super Intelligence Research Institute
Meta is gearing up for a major AI update, working to improve image, video, and text models for release in early 2026. According to the Wall Street Journal.
This is an initiative of Meta's Superintelligence Institute. led by Alexander WangHe is a co-founder of Scale AI and is currently leading the company's most ambitious AI plan to date.
Also, the meta is at the same time Image and video model development This is code-named “Mango” and the text-focused model is called “Avocado”. The plans were reportedly revealed during an internal Q&A session in which Wang and Meta's chief product officer Chris Cox explained how these models will shape the company's future AI strategy.
Beyond chatbots
The goal applies not only to chatbots, but also to “world models.”Mr. Wang says: Meta's new text-based model aims to deliver a noticeable improvement in coding skills, a key aspect where the likes of OpenAI and Anthropic are leading the way.
At the same time, tech giants are making further progress through the development of “world models” that can not only process visual data and reason about it, but also plan and act in all situations without being explicitly trained.
This move marks a shift from simple assistants to more independent, multimodal systems, AI that not only reacts but also interprets and predicts. If they can pull it off, Mango and Avocado could be Meta's first serious step into the world of general purpose intelligence.
catch up with AI
The company's AI division experienced disruption, which led Meta to refocus its efforts. They faced setbacks from the likes of OpenAI, Google, and Anthropic, which resulted in reorganizations, management changes, and aggressive hiring from other labs. Still, the transition wasn't smooth.
Many of the top researchers hired by Meta's superintelligence institute have retired; their resignation It is a cause of instability within the organization. The uncertainty of the situation was further exacerbated last month by Meta's Chief AI Scientist Yann LeCun: decided to quit the company To start your own venture.
Although Meta's technology and staff are among the best, instability issues are always present.
There has been no breakthrough in AI so far.
Meta has a huge user base, but none of its AI products are clearly market leaders. The number of users of Meta AI is primarily due to its distribution rather than the demand for it. This is because the Assistant is integrated into the search bar for Facebook, Instagram, and WhatsApp.
The size of the user base is huge, but that doesn't necessarily mean the company is a technology leader. This situation makes the first model released by SuperIntelligence Lab even more difficult. Mango and avocado are not only new products, but also a test of the company's credibility.
conclusion
According to Meta's 2026 AI roadmap, mango and avocado are likely to see significant reductions, or simply the company's AI restructuring phase. This suggests that the company recognizes lost ground, but still has considerable confidence in its ability to change the story.
With a new management team, ambitious technology goals, and an embedded audience of billions of users, the benefits are huge. Whether Mango and Avocado become breakthroughs or just notes will be the deciding factor in whether Meta regains its AI momentum.
