Meta $META ( ▲ 5.68% ) is trying to reset the narrative after the messy Llama 4 rollout.

At the World Economic Forum in Davos, CTO Andrew Bosworth said Meta’s Superintelligence Labs team delivered new AI models internally this month, and according to him, they’re “very good.”

That matters because Llama 4 wasn’t exactly a clean win. The last release got dragged by delays and accusations that some benchmark results were rigged, which put a cloud over Meta’s open model momentum.

What Meta is likely building

Bosworth didn’t say what the new models are, but The Wall Street Journal has previously reported Meta is working on:

  • Avocado: a large language model

  • Mango: an AI image and video model

So the implication is Meta’s next wave may be broader than “just another Llama,” with multimodal upgrades baked in.

Meta basically wants to prove this wasn’t a one-off stumble.

If these models ship publicly and actually hold up, the Llama 4 controversy becomes a footnote.

If not, it becomes a pattern.

Reply

or to participate

Keep Reading

No posts found