First AI models from Meta's Superintelligence Lab out internally
What's the story
Meta Platforms's newly established artificial intelligence (AI) lab, the Meta Superintelligence Labs, has successfully delivered its first set of high-profile AI models. The announcement was made by Andrew Bosworth, the company's Chief Technology Officer (CTO), during a press briefing at the World Economic Forum's annual meeting in Davos. Bosworth said that the models created by his team in just six months are "very good."
AI projects
Meta's AI models: Avocado and Mango
In December, media outlets reported that Meta was developing a text AI model codenamed "Avocado," which is expected to launch in the first quarter of this year. The company is also working on an image and video-focused model codenamed "Mango." However, Bosworth did not disclose which specific models were delivered internally by his team at the press briefing.
Competitive landscape
AI efforts amid industry competition
Meta's AI endeavors have been closely monitored, especially after Mark Zuckerberg's decision to revamp the company's AI leadership and establish a new lab. The tech giant had also been criticized for its Llama 4 model performance, as competitors like Alphabet's Google gained momentum in the race for transformative and lucrative AI technologies.
Development insights
Bosworth on AI development cycle
Bosworth emphasized that the technology is still a work in progress. He explained that there's a lot of post-training work to be done for AI, to make models usable internally and by consumers. Despite these challenges, he said Meta is starting to see favorable returns from its big gambits in 2025, which he described as a "tremendously chaotic year" for building its lab and infrastructure.
Future predictions
AI trends for 2026 and 2027
Looking ahead, Bosworth predicted that consumer AI trends would firm up in 2026 and 2027. He said recent advances had already produced models capable of answering everyday questions. However, he acknowledged that technological development could yield better results for more complex queries. The next two years are crucial for bringing consumer products to market, he added.