Meta’s New AI Team Delivered First Major Models Internally, CTO Andrew Bosworth Confirms

Meta’s new AI organization has unblocked its first gigantic AI models for internal usage, as per the declaration made by Andrew Bosworth, the Chief Technology Officer. He presented the news at the World Economic Forum in Davos, where he indicated that the group has achieved a lot during the less than six months of its work.

The group works under the name of Meta Superintelligence Labs, a new AI unit that has been established under the direction of CEO Mark Zuckerberg’s grand plan to make Meta the leader of the AI race. Meta’s AI policy has been a subject of constant scrutiny, particularly after internal discontent with previous AI performance and growing competition.

In his alias comments, Bosworth did not mention the exact models. Nonetheless, Reuters points out that previous reports alluded to two significant internal undertakings: a text-oriented model said to be codenamed Avocado and an image or video model said to be codenamed Mango. Such information implies that the team is developing skills in both language and visual AI, which are the primary areas of modern consumer applications.

Bosworth has underlined that model training is just the first step toward the goal. The model needs a lot of “post-training” work done to it before it can be used in real products, even if it is already built. This work includes refinement, testing, safety layers, performance upgrades, and system integration of the user’s touch points.

Meta has put tons of money into building up its AI infrastructure, and according to Bosworth, the company started to feel the effects of its 2025 infrastructure spending positively. He speculated that the years 2026 and 2027 would be the main years when those investments will bring forth consumer-friendly AI releases.

In this regard, Reuters has also pointed out Meta’s Ray-Ban smart glasses initiative. The firm has halted the international rollout of certain Ray-Ban Display glasses, concentrating on U.S. demand while working on larger AI product plans.

This news is significant because it indicates that Meta’s AI reboot is yielding results quicker than what was the general expectation. The provision of internal models signifies that Meta is regaining momentum and getting ready for the next product cycle where AI features are expected to be the main factor driving user engagement across social apps, hardware, and new interfaces.

For Meta, this is not just about models being shipped. It is about showcasing that the new team structure and the new strategy are capable of bringing about AI advancements that are on par with the best in the industry.