Meta Platforms has achieved a significant milestone in its artificial intelligence development, with its newly established Meta Superintelligence Labs delivering its first high-profile AI models internally this month. Chief Technology Officer Andrew Bosworth confirmed the development during a press briefing at the World Economic Forum’s annual meeting in Davos on Wednesday.
The technology executive revealed that the team, formed just six months ago as part of CEO Mark Zuckerberg’s strategic reorganization of Meta’s AI leadership structure, has already produced models demonstrating substantial promise. While Bosworth did not specify which particular models were delivered, previous media reports indicated Meta was developing a text-based AI system codenamed ‘Avocado’ scheduled for first-quarter release, alongside an image and video-focused model known as ‘Mango’.
This development follows Zuckerberg’s aggressive moves to reposition Meta in the intensely competitive AI landscape, including leadership restructuring, establishing specialized labs, and recruiting top talent with exceptionally generous compensation packages. The initiatives represent Meta’s strategic response to criticism surrounding the performance of its Llama 4 model and the significant advancements made by competitors including Alphabet’s Google.
Bosworth emphasized that while the initial models show impressive capability, the technology remains unfinished. He detailed the extensive post-training work required to refine AI systems before they become usable both internally and for consumer applications. Despite these ongoing development challenges, the CTO indicated that Meta’s substantial investments and strategic gambits initiated throughout 2025 are beginning to yield favorable returns, positioning the company more competitively in the race for transformative AI technology.