Meta’s New AI Lab Unveils First Internal Models in Early Step Toward Superintelligence
Meta Platforms has unveiled the first high-profile artificial intelligence models developed by its newly formed AI laboratory, marking an early milestone in the company’s push to compete in the rapidly evolving AI industry.
Speaking during a press briefing on the sidelines of the World Economic Forum’s annual meeting in Davos, Meta Chief Technology Officer Andrew Bosworth confirmed that the company’s Meta Superintelligence Labs has successfully delivered its first internal AI models this month.
According to Bosworth, the team—formed just last year—has already produced promising results despite being in the early stages of development.
“They’re basically six months into the work, not quite even,” Bosworth said. “The models are very good.”
While Bosworth did not reveal specific details about the models released internally, earlier reports indicated that Meta has been developing several AI systems under code names. These include “Avocado,” a text-based AI model expected to launch in the first quarter, and “Mango,” an AI model focused on image and video generation.
Meta’s new AI lab is being closely watched across the technology industry following significant leadership changes initiated by CEO Mark Zuckerberg. The company reorganized its AI efforts, established the dedicated research lab, and aggressively recruited top AI talent with highly competitive compensation packages.
The move comes as Meta faces intense competition from other technology giants such as Google, which has recently gained momentum in the global race to develop advanced AI systems.
Meta has also faced criticism over the performance of its earlier Llama 4 model, raising pressure on the company to strengthen its capabilities in artificial intelligence.
Bosworth acknowledged that the newly developed models are still undergoing further refinement before they can be widely deployed.
“There’s a tremendous amount of work to do post-training,” he said, explaining that AI systems require significant testing and optimization before they become usable both internally and for consumers.
Despite the challenges, Bosworth noted that Meta’s strategic investments and restructuring efforts in 2025 are beginning to show encouraging progress.
As the competition to build powerful and transformative AI systems intensifies, Meta’s early results from its Superintelligence Labs signal that the company is accelerating its efforts to stay competitive in the global AI race.
