WORLD NEWS
Meta Platforms (META.O) has unveiled the first high-profile artificial intelligence models developed by its newly formed Meta Superintelligence Labs, the company’s Chief Technology Officer Andrew Bosworth revealed on Wednesday.
Speaking on the sidelines of the World Economic Forum (WEF) 2026, Bosworth said the models, created by the lab established last year, “showed a lot of promise” despite being only six months into development.
“They’re basically six months into the work, not quite even,” Bosworth said, describing the internal models as “very good.” He did not confirm which specific models had been delivered but recent reports indicate Meta is working on a text AI model codenamed Avocado and an image/video-focused model codenamed Mango, both slated for first-quarter launches.
The new AI efforts come after Meta faced criticism over the performance of its Llama 4 model, while competitors like Alphabet’s Google have gained momentum in the AI race. Bosworth noted that post-training work is critical for AI, emphasizing that models must be made usable both internally and for consumers.
Meta has also made strides in consumer-facing AI, including AI-equipped Ray-Ban Display glasses, whose international rollout was recently paused to prioritize U.S. orders. Bosworth called 2025 a “tremendously chaotic year” as the company built its lab, infrastructure, and computing power.
Looking ahead, Bosworth said 2026 and 2027 will be key years for consumer AI adoption, as recent advances already allow models to answer everyday questions and can be improved for more complex tasks.
Meta’s aggressive push into AI, including leadership reshuffles and talent acquisition with “sky-high offers,” underscores its aim to compete at the forefront of the technology frontier.