Meta has announced the internal delivery of new AI models this month, claiming they are “very good.” This comes after the release of Llama 4, which faced delays and scrutiny over alleged rigged benchmarks. Andrew Bosworth, the company’s CTO, shared this update at the World Economic Forum, highlighting the progress made by the Superintelligence Labs team. While Bosworth didn’t disclose specific details about the new models, reports from The Wall Street Journal indicate that Meta is developing a large language model and two AI models for image and video, known as Avocado and Mango. The shift suggests Meta is focusing on improving its AI capabilities following past setbacks and is eager to regain credibility in the competitive AI landscape.