Meta’s New AI Lab Delivers First Internal Models: CTO Says

Meta Platforms’ new artificial intelligence lab delivered its first major AI models internally this month, the company’s chief technology officer said on Wednesday.

Speaking at a press briefing on the sidelines of the World Economic Forum annual meeting in Davos, CTO Andrew Bosworth said the models developed by Meta Superintelligence Labs—formed last year—were showing strong early results.

“They’re basically six months into the work, not quite even,” Bosworth said, describing the models as “very good.”

Media reports in December said Meta was developing a text-based AI model codenamed Avocado, expected to launch in the first quarter, along with an image- and video-focused model called Mango. Bosworth, known internally as “Boz,” did not specify which models had been delivered internally.

Meta’s AI push is being closely watched following major moves by CEO Mark Zuckerberg to overhaul the company’s AI leadership, create a new lab, and recruit top talent with lucrative compensation packages. The company is seeking to regain ground in an increasingly competitive AI race.

Meta has faced criticism over the performance of its Llama 4 model, as rivals such as Google and other tech firms gained momentum in developing advanced and commercially viable AI systems.

Bosworth said the work was far from complete.

“There’s a tremendous amount of work to do post-training,” he said, noting that models must still be refined to make them usable internally and by consumers.

Still, he added that Meta’s large AI investments made in 2025 were beginning to show positive results.