According to AIbase, Meta is launching an unprecedented "full-scale counterattack" in the field of artificial intelligence. Leading this effort is Alexandr Wang (former founder of Scale AI, now leading Meta's Super Intelligence Lab), who plans to release a series of next-generation intelligent models with highly "tropical" code names in the first half of 2026.

Facebook Metaverse meta

The focus of this release includes the multi-modal model named Mango, designed to unify the generation and understanding of images and videos. At the same time, Meta is also secretly developing the next-generation large language model Avocado (Avocado). According to insiders, Avocado's core goal is to achieve a generational leap in coding capabilities and to explore the "world model"—that is, enabling AI to build an understanding of the real physical environment through visual information.

In order to regain technological dominance, Meta CEO Mark Zuckerberg personally launched a talent "competition," luring more than 20 core researchers from OpenAI. This move is aimed at responding to the increasingly fierce competition in multi-modal technology:

  • Google has achieved explosive growth in user numbers with its Nano Banana image creation tool (based on Gemini2.5Flash Image), with monthly active users exceeding 650 million.

  • OpenAI entered a "red alert" state after the release of Gemini3, guarding its moat with Sora and the upgraded ChatGPT Images.

  • Meta previously collaborated with Midjourney to launch the Vibes video generator, and the upcoming 2026 product series is seen as a key step in realizing its vision of "personal super intelligence."