Meta Platforms (META) Chief Technology Officer Andrew Bosworth stated on Wednesday that the company’s newly formed AI laboratory delivered its first batch of high-profile AI models internally this month.

Speaking at a press briefing during the World Economic Forum annual meeting in Davos, Bosworth noted that the models built by Meta’s “Superintelligence Lab”—a team assembled last year—have demonstrated “tremendous potential.”

“They’ve essentially been at this for six months, maybe even less,” Bosworth said, adding that the performance of the team’s AI models has been “truly excellent.”

In December, media reports indicated that Meta was developing a text-based AI model codenamed Avocado, planned for a first-quarter release, alongside an image and video model codenamed Mango. Bosworth did not explicitly specify which models were included in this internal delivery.

Meta’s efforts have been under close scrutiny following CEO Mark Zuckerberg’s aggressive moves to overhaul AI leadership, establish new labs, and lure top talent with astronomical compensation packages. Zuckerberg aims to secure a victory for the company in this fiercely competitive technological frontier.

Previously, Meta Platforms (META) faced criticism regarding the performance of its Llama 4 model. Meanwhile, competitors such as Google have gained an early lead in this transformative and highly lucrative AI race.

Bosworth clarified that the technology is not yet finalized.

Commenting on the general development cycle, Bosworth said, “In the field of AI, there is a massive amount of ‘post-training’ work to be done to ensure the delivered models are usable for both internal purposes and for consumers.”

However, he noted that Meta’s significant “big bets” for 2025 are already beginning to show promising returns.