Meta Is Developing a New AI Model for Image and Video Generation: Code-Named ‘Mango’

In a significant step to strengthen its position in the rapidly evolving generative AI landscape, Meta Platforms is reportedly developing a new artificial intelligence model focused on creating high-quality images and videos. The project, internally code-named “Mango, was revealed during an internal company Q&A session on December 18, 2025, according to sources familiar with the discussions.

Meta’s Chief AI Officer, Alexandr Wang, and Chief Product Officer, Chris Cox, shared details about the upcoming model with employees. Mango is part of a broader push by Meta’s specialized AI division, Meta Superintelligence Labs (MSL), which was established earlier in 2025 following a significant restructuring of the company’s AI efforts. The model is expected to launch in the first half of 2026, alongside a separate next-generation text-based large language model code-named “Avocado.

Competing in the Generative Media Race

The development of Mango comes at a time when image and video generation has become one of the most competitive battlegrounds in AI. Rivals such as OpenAI (with Sora for video and advanced image tools in ChatGPT) and Google (with Gemini and its image tool Nano Banana) have already released competent multimodal models that attract massive user engagement.

Meta has been active in this space for years, with earlier efforts including:

• Make-A-Scene (text-to-image and video tools from 2022),
• Llama Image foundation models,
• Movie Gen (a research breakthrough for video generation and editing),
• Vibes (an AI video generator launched in collaboration with Midjourney in September 2025).

Mango aims to build on these foundations, offering advanced capabilities such as multimodal generation (combining text, images, and video inputs), realistic motion handling, and integration with Meta’s ecosystem of apps, including Instagram, Facebook, and WhatsApp. Reports suggest it could power enhanced features for creators, including better Reels, Stories, AR filters, and generative advertising tools.

A Shift in Strategy?

The Mango project may signal a subtle shift for Meta, which has traditionally favored open-source models like the Llama family. While details remain limited, some analysts speculate that Mango could follow a more proprietary approach to compete directly with closed-source offerings from OpenAI and Google. This aligns with Meta’s aggressive talent acquisition—CEO Mark Zuckerberg personally recruited over 20 researchers from OpenAI—and its $14 billion+ investment in Scale AI to bolster MSL.

Wang also mentioned early exploration of “world models, AI systems that learn physical realities from visual data to enable more realistic simulations—far beyond simple text-to-video generation.

What It Means for Users and the Industry

For everyday users, Mango could bring more intuitive and powerful AI tools directly into Meta’s apps, reducing reliance on third-party generators and boosting engagement. For creators and advertisers, it promises faster workflows, lower production costs, and innovative formats.

However, the timeline is still preliminary, and Meta has not yet confirmed public details. The company continues to emphasize responsible AI development, including safeguards against misuse.

As the AI arms race intensifies, Mango represents Meta’s bid to catch up and leapfrog competitors in multimodal generation. Stay tuned for official announcements in 2026—Meta’s next chapter in generative AI could reshape how billions create and consume visual content.