Key Moments
- Meta’s new AI lab delivered its first major models internally this month, the company’s CTO said.
- The models were built by the Meta Superintelligence Labs team formed last year and show “a lot of promise.”
- Meta’s AI push remains under scrutiny after leadership changes, the new lab launch, and aggressive talent hiring.
Internal Debut of New AI Models
DAVOS, Switzerland, Jan 21 (Reuters) – Meta Platforms’ new AI lab rolled out its first major models for internal use this month, CTO Andrew Bosworth said.
He spoke during a press briefing at the World Economic Forum in Davos. Bosworth said the Meta Superintelligence Labs team, formed last year, is already producing strong results.
“They’re basically six months into the work, not quite even,” Bosworth said. He added that the models are “very good.”
Early Progress From Meta Superintelligence Labs
Bosworth said the team advanced quickly and began delivering high-profile models within about six months. He also noted that the models in testing are promising, but he did not share technical details.
Code-Named Projects and Unspecified Deliveries
Media outlets reported in December that Meta was developing a text model code-named Avocado and an image/video model called Mango. However, Bosworth did not confirm whether the internal models match those projects.
| Model Code Name | Focus | Timing | CTO Comment |
|---|---|---|---|
| Avocado | Text AI | First-quarter launch (reported) | Not confirmed |
| Mango | Image/video AI | Reported alongside Avocado | Not confirmed |
Strategic AI Push Under Zuckerberg
Meta’s progress draws close attention after CEO Mark Zuckerberg reshaped the company’s AI strategy. He reorganized AI leadership, launched the new lab, and recruited top talent with generous offers.
Overall, Meta aims to strengthen its position in a highly competitive AI frontier. Investors and analysts will watch closely to see how these internal advances translate into future products.





