Meta is quietly working on a new artificial intelligence system that can create images and videos, according to recent reports. The company plans to release this new model in the first half of 2026. Inside Meta, the project is known by the code name Mango. Alongside it, Meta is also building a text based model called Avocado, which is meant to be better at coding and tool use.
This move shows that Meta is taking the AI race very seriously. Other companies like OpenAI, Google, and Anthropic have been moving fast. Meta does not want to be left behind. With Mango and Avocado, Meta is trying to build stronger AI products that can compete at the highest level.
In this post, we will break down what Meta is building, why it matters, and what it could mean for users, creators, and the future of AI. Everything is explained in very simple words.
What Meta is working on
Meta is developing a new AI model that can understand and create images and videos. This model is called Mango inside the company. It is being built by Meta’s superintelligence lab, which is led by Alexandr Wang, the co-founder of Scale AI.
Mango is not just about making pretty pictures. The goal is to create high quality images and long videos that make sense from start to finish. The system is designed to understand how the real world works, things like objects, movement, light, and space.
Meta is also building another model called Avocado. This one focuses on text, especially coding and planning tasks. Avocado is meant to help the system think, plan actions, and use tools more effectively.
Together, Mango and Avocado could work as a team. One model handles thinking and planning, while the other handles images and videos.
Why Meta needs this
Meta has fallen behind in the AI race. Companies like OpenAI have gained attention with tools like ChatGPT and video models like Sora. Google has also shown strong image and video systems. Meta, on the other hand, does not yet have a clear AI product that people love on its own.
Meta’s AI assistant is used by many people, but mostly because it is placed inside Facebook, Instagram, and WhatsApp. People do not go to Meta AI because it is the best. They use it because it is already there.
This is a big problem for Meta. If people start choosing other AI tools on purpose, Meta could lose its influence over time. That is why the company is putting a lot of pressure on its AI teams to deliver something strong.
The first models coming out of the superintelligence lab have a lot riding on them.
What makes Mango different
Mango is designed to be a world model. This means it does not just guess what a video should look like. It tries to understand how the world works.
For example, if a person walks across a room, the system should keep the same person looking the same from start to end. Lighting should stay realistic. Objects should not change shape for no reason. Gravity should work the way it does in real life.
This is one of the hardest problems in AI video today. Many video tools can make short clips, but they struggle with longer videos that stay consistent. Meta wants Mango to handle long videos with better control and realism.
If Meta succeeds, this could be useful for creators, advertisers, and everyday users.
When it could launch
According to reports, Meta plans to release Mango in the first half of 2026. This is an aggressive timeline, but it is possible if Meta continues to invest heavily in computing power.
Meta has been buying large numbers of powerful chips, including Nvidia H100 GPUs. These chips are needed to train and run advanced AI models. The company wants to support billions of requests across its apps, so it needs massive data centers and fast systems.
Releasing Mango on time will depend on how fast Meta can build and scale this infrastructure.
How this could be used
If Mango and Avocado work as planned, Meta could use them in many ways.
Creators could make videos, edit clips, or design images directly inside Instagram or Facebook. Advertisers could generate video ads faster and localize them for different regions. Businesses could create product videos without hiring large teams.
Meta could also automate parts of content creation. For example, turning a long video into short highlight clips or creating many versions of the same content for different audiences.
This could change how content is made on Meta platforms.
The big challenges ahead
There are still major challenges. Training video models requires huge amounts of high quality data. Meta has access to a lot of user content, but this raises privacy and legal concerns, especially in places like Europe.
Safety is another big issue. Realistic video AI can be used to create deepfakes. These can harm people, spread lies, or affect elections. Meta will need strong safety systems, watermarking, and clear rules to prevent misuse.
The public will judge Mango not just by how good the videos look, but by how safe the system is.
The Bottom Line
Meta wants a win. Right now, the company relies on its social apps to stay relevant. A strong AI model could help Meta stay competitive for the next decade.
If Mango succeeds, Meta could become a leader in image and video AI. If it fails, other companies may take that position instead.
Next year will be very important. Watch for early demos, research updates, and signs of how close Meta is to making Mango a real product.One thing is clear. Meta is betting big on AI, and 2026 could be a turning point for the company.
Also Read:OpenAI is reportedly trying to raise $100B at an $830B valuation
